My research group works on two different areas
• Spatial Audio and Computational Sound Scene Analysis:
I have been fascinated by the way humans effortlessly use spatial sound for making sense of the environment. A desire to understand the mechanisms used and to provide machines with the same capabilities has been a driver
in this research. The use of computing, mathematical physics and experiments have characterized this work.

• Large Scale Scientific Computing
In scientific computing I focus on very large problems using approximation algorithms and parallel computing. The work has been driven by applications in audio, speaker ID, vision, physics, and machine learning. The focus here is on extending algorithms based on the fast-multipole method, and the use of data-parallel architectures to achieve orders of magnitude speed-ups.

I am interested broadly in computer vision and multimedia processing for human-centric interactive applications involving home-based healthcare, smart environments, and educational applications.

We have many projects that span multi-agent AI, robotics, and biology. Please see our website for more information.

My areas of expertise are in Natural Language Processing and Computational Linguistics. The long term goal of my research is to develop methods and systems that automatically discover the
meaning of written text and understand its content. To that end, I am developing techniques that
(1) cluster short text snippets based on the similarity of their content, (2) discover word meanings in large corpora of text, (3) assign meanings from a dictionary to words in running text, and (4) measure the semantic similarity and relatedness between concepts. I place a high priority on releasing and maintaining free open source software packages to enable others to reproduce and extend our work.

My research interests are in the area of computer graphics, particularly in data-driven materials, virtual humans, and computational photography and display.

Hierarchical graphical models for object and activity recognition; Max independent set and max clique formulations of vision problems, such as image matching and tracking

Intra-die process variations in nm technology nodes pose significant challenges to robust design practices. Geometric variations along with random dopant fluctuation effects have had significant impact on Logic/Memory functionality/yield. The inaccuracies in the models and variabilities in the process are more pronounced and key is to understand the variability effects. Innovative analysis techniques as well as methodologies are needed to counteract the variability issues. Our team is heavily engaged in this activity.

Smart statistical and numerical techniques are key for understanding the variability in VLSI circuits. These techniques need to be highly efficient to be used by designers. New algorithms to achieve speed-up in these techniques are essential. Extension of these smart algorithms beyond the VLSI domain would be critical to achieve accurate and computationally practical would benefit such fields (e.g application to medical field). We are exploring new algorithms and their applications to fields other than VLSI.

As the VLSI circuits scale power and performance remain the top issues. Developing efficient circuit techniques based on emerging devices (FinFet derived, nano-devices) needs extra attention.
We are involved in fabrication of such circuits. Also we are engaged in modeling of 3-D structures and the thermal management of multilevel domain.

Scaling and manufacturing of alternate memories such as MRAM and PCM would help to replace DRAM. Variety of activities from modeling to fabrication are explored in our team

My my main interests involve social systems informatics. The this interdisciplinary program in Social Systems Informatics is directed towards advancing and disseminating computational approaches involving the measurement, modeling, and testing of how social interactions affect a broad range of human outcomes. These social interactions range from face-to-face, media communications, or other social actions and are often massive in size. Key outcomes involve major characteristics of our lives related to education, health, work, and play; attitudes, behaviors, performances in these areas are not simply products of individual factors but are fundamentally shaped by the social contexts and interactions in which people live. Computationally, social systems informatics relies on methods in system science, including social network analysis, agent based modeling, data mining, intelligent data analysis, and biostatistical methods, to study how individual and social system interactions affect health and other outcomes.

Research in the Computer Vision Laboratory at Columbia University is focused on three broad areas, namely, the creation of novel vision sensors, the design of physics based models for vision, and the development of algorithms for scene interpretation. Research projects are motivated by applications in the fields of digital imaging, computer graphics, human-machine interfaces and robotics.

Smart Visualization: One picture is worth a thousand words. For thousands of years, people have been using information graphics—visual representation of data—to comprehend and analyze information. However, creating high-quality visualization is a daunting task especially for ordinary people who are neither graphic artists nor computer scientists. To democratize the use of visualization, I am interested in automated visualization design and tailoring the visual responses to highly dynamic user interaction situations and unanticipated information. More recently, I have been working in the area of interactive visual text analytics, which combines state-of-the-art text analytics with novel interactive visualization to empower average business users to analyze massive amounts of textual data and facilitate their decision making (e.g., making a purchase decision based on the visual text analysis of extensive consumer reviews).

Mixed-initiative Human-Computer Interaction (HCI): I believe that the future of HCI is to facilitate the development of a man-computer symbiosis where both humans and machines can leverage their strengths and avoid their weaknesses. I am especially interested in the development of mixed-initiative intelligent information systems, where the humans and computers work together collaboratively to facilitate information seeking an analysis. To enable computers to take sensible initiatives and push this class of systems to main stream applications, I am particularly interested in developing novel and practical computational approaches to the problem. Moreover, I am interested in exploring new interaction paradigms where users can interact with complex system responses (e.g., system-derived text summarization results) and the use of interactive machine learning in support of an adaptive, mixed-initiative human-computer interaction, where both humans and computers can learn from each other.

Opportunistic Social Computing: The use of social software (e.g., social networking, micro-blogging, and online forums) has penetrated the masses. I am very much interested in finding out how such phenomena will change our daily lives as well as its long-term impact on our world. In particular, I am especially interested in how social computing will bring us opportunistic information and collaboration partners whenever we need them but without subjecting ourselves to “constant availability and instant intimacy” as we do today. To achieve this goal, I believe there are fundamental research issues to be addressed.They include but not limited to:

•Understanding, modeling, and automatically deriving social profile of a person, a community, or an organization based on their digital footprints (i.e., online behavior);
•Use of the derived social profiles to objectively reveal key characters of individuals and organizations, assess community/organization dynamics, value, and risks, predict the development or growth of individuals, communities and organizations, and help establish opportunistic collaborations among individuals and organizations;
•Monitoring social channels (e.g., facebook and twitter) and detecting most valuable channel(s) for extracting social intelligence (e.g., knowledge about car repair or the consumer complains/needs); Mining of social messages to distill insight (information or people) for opportunistic information sharing (e.g., sharing the extracted consumer complains), knowledge acquisition (e.g., asking target audience to voice their problems and suggest their solutions), and crowd-sourced problem solving (e.g., social Q&A). See my webpage for more info.

I am directing Data Mining Group in the Center for Computational Science at the University of Miami. Overall goals of the group is to develop methods for for understanding, spatial/temporal, heterogeneous data.

The popularity of web-based mapping services such as Google Earth/Maps
and Microsoft Virtual Earth (Bing), as well as the increasing use of
smartphones, has led to an increasing awareness of the importance of
location data and its incorporation into both web-based search
applications and the databases that support them.

Location data is a subset of the broader area of spatial and
multidimensional data, and my interests lie in developing efficient
representations and algorithms for its use in applications in the
computer graphics, databases, computer vision, geographic information
systems (GIS), image processing, high dimensional data, and search
domains. My work is based on the observation that these
representations invariably reduce to finding ways to sort the data
with the caveat that sorting is a linear process, having a reference
point, which yields an explicit order. The drawback here is that in 2
and higher dimensions, the data must be resorted when the reference
point is changed. This is not the case with methods that use implicit
sorts such as those that sort objects with respect to the space that
they occupy (e.g., quadtrees, octrees, R-trees, etc.). I characterize
the problem as one of “sorting in space”. I am interested in
developing algorithms for problems in the above domains that are based
on the notion of sorting. Some examples include finding nearest
neighbors and shortest paths in spatial networks where the distance is
along the edges of the network rather than as “the crow flies”,
as well as finding similar trajectories.

Although location data is usually specified geometrically, many of
today’s web-based applications increasingly find it being specified
textually as it corresponds to place names. I am interested in
techniques for identifying such data as well as resolving it to the
right geometric location as these specifications are often ambiguous.
I have a number of projects that try to make use of such data for
searching databases such as those in the deep web as well as
spreadsheets, news, tweets, to name a few, and in mobile applications.

Bo Begole is a Principal Scientist and Manager of the Ubiquitous Computing Area at the Palo Alto Research Center. He is an applied computer scientiest who aims to create novel systems and capabilities for end-user applications. His past work includes systems that provide synchronous collaboration of single-user applications, computer-mediated communication, distributed interpersonal awareness, sensor-based interruptibility detection, presence modeling and prediction, media device interoperability and control, reducing information overload and context-aware mobile systems.

Bo’s unifying research interest is in the area of Behavior Modeling — Using sensor and data inputs, along with inference algorithms, to identify and predict an individual’s current activity and interest in particular information. He likes to work with people who have broad interests and creative approaches to solutions.

Bo received a B.S. in 1992 in Mathematics from Virginia Commonwealth University, an M.S. in 1994 and a Ph.D. in 1998 in Computer Science from Virginia Tech. Prior to his studies, Bo served in the US Army as an Arabic language translator specializing in Egyptian, Libyan and Iraqi dialects. Bo was recently a co-Chair of the 2008 conference on Computer-Supported Cooperative Work (CSCW 2008) and was the Technical Program chair of CHI 2011.

My current research is focused on basic and applied research in computer vision, image understanding, pattern recognition, and quantitative microscopy. The mission of our group has been to develop novel methods and systems for image and scene analysis that are capable of making human-like decisions. Our passion has been in the pursuit of scientific excellence and innovative engineering, enabling pragmatic solutions to problems of societal impact. Current research initiatives of the group are in wide-area video surveillance for behavioral analysis and biometrics and biomedical image analysis for morphological characterization with applications in medicine.

Our work focuses on how a robot can autonomously learn the foundational domains of higher-level knowledge — space, objects, actions, affordances, and so on — from unguided experience with the “blooming, buzzing confusion” of pixel-level sensorimotor interaction with an unknown world.

We have done work on learning about sensors, constructing models of large-scale space, abstracting foreground objects from the relatively static background, and building a hierarchy of progressively higher-level actions starting from simple motor signals.

Some of our work starts at a higher level, developing a robot that can serve a person as a mobility aide. It needs to build its own cognitive map of the environment from its experiences, but it also needs to learn to communicate effectively with the human driver, who is the ultimate source of authority in its world.

DEMO focuses on origins and principles of organization for complex systems. If we can create chain reactions in complexity with a universal computational substrate, then certain hard problems should become easier to solve.

Co-Evolutionary Learning

Most learning takes place as optimization of a fixed environment or fitness function. This requires that the learner be “pre-adapted” to that environment in order to learn anything, or that the environment be “gradient engineered” for the particular learning mechanism. This inductive bias usually makes the researcher the primary cause of learning. We focus on dynamically-changing learning environments, often composed of competing learners, where the complexity of the task gradually and automatically increases without human intervention. We work in games, language learning, and computational optimization tasks.

Evolutionary Machines

The long term goal is the co-evolution of machines and their brains, first in simulation, then, through advanced computer-aided manufacturing, into actual hardware. Our initial experiments simulated LEGO structures and evolved groups of simulated robot agents. The GOLEM project through 5 generations evolved designs which could be built into reality using automated manufacture. The frontier is to co-evolve form and formation, to evolve manufacturing plans which can overcome noise and errors.

Theory and algorithms for machine learning. Topics include:
- Design and analysis of algorithms for clustering data streams
- Climate Informatics: a new subfield on machine learning for climate science
- Online learning, and learning from data streams (including in semi- or unsupervised settings)
- Privacy-preserving machine learning
- Active learning

Carlos is interested in developing efficient algorithms and methods for designing, analyzing and controlling complex real-world systems. Our application domains have been typified by sensor networks, where limited resources, unreliable communication, failures, and large problem scale are prevalent. These characteristics require us to develop theoretically-founded algorithms, to unify methods from statistics, optimization theory and machine learning, and to exploit problem-specific structure. Moving forward, our focus will be on designing methods for managing information overload, and tackling the challenges of machine learning over parallel architectures.

Building behavioral/content based user profiles and recommender systems

We are seeking highly qualified research fellows in content analysis, text mining, machine learning, recommendation systems, and data mining. Example areas include (but are not limited to) classification, similarity analysis, clustering, feature selection, search, text mining, personalization, user modeling, collaborative filtering, novel information processing and extraction methods, and related areas. The candidate will make contributions in a strong team of diverse backgrounds (computer science, information management, statistics, econometrics, mathematical optimization, and marketing science). The charter of this team is to generate knowledge and build technologies to spur the growth of HP’s businesses.

Imre Solti is an Assistant Professor in the Division of Biomedical Informatics, Department of Pediatrics at Cincinnati Children’s Hospital Medical Center at Cincinnati, Ohio. He received his M.D. in 1992 from Albert Szent-Gyorgyi Medical School, Szeged, Hungary, his Ph.D. in Health Services Organization and Research from the Medical College of Virginia, in 2006, and his M.A. in Computational Linguistics from University of Washington, Seattle in 2011. He received a K99 Career Development Award from the National Library of Medicine (NIH) in 2009. He is a current NIH R00 Pathway to Independence Award recipient (2010). His recent research interests are based on information extraction from electronic health records to automate the clinical trial eligibility screening process, develop clinical Natural Language Processing and Clinical Decision Support Systems for patient safety, and predictive models for clinical outcome.

HP Labs Immersive 3D

HP Labs is conducting research using large display walls, 2D and 3D. HP has multiple product and services offerings in the so-called “Big Walls” area ranging from digital signage to large telepresence facilities (such as HP’s Halo). Our research extends this by following two main, intertwined branches. The first examines a new style of 3D entertainment, and the other targets commercial and industrial uses for 3D enabled operations centers, brainstorm or war rooms, and control centers. We are conducting research in:
• Computer vision, image processing, and recognition,
• Robust multi-imager and multi-camera modeling and calibration,
• 3D data visualization,
• High performance computation, transport, and imaging on hardware accelerated platforms (CPUs and/or GPUs),
• Novel human-big wall interaction modalities across heterogenous systems, and
• Local and remote collaboration technologies.

We have publicly shown our Immersive 3D Entertainment efforts with events at the 2010 Sundance Film Festival and the Earth Wind and Fire Concert at 2011 CES in Las Vegas (links listed below). HP’s goal is to experiment with non-standard aspect ratios (e.g. 3:1 for concert stage, 5.5:1 for basketball court) to allow the audience to experience the event as though you are seated at the best seat in the house. This is a break from the traditional movie view from the director’s eyes and allows the audience to “look around” and soak up the scene which works so well in 3D. To do this, we are experimenting with large multi-projector 3D displays (Pluribus), multi-imager camera capture (Herodion), and advanced digital image pipelines (Pericles). We conduct in-situ laboratory studies at entertainment events such as concerts, sports events, fashion shows, and other gaming.

We have built an experimental operations center with multiple 2D and 3D walls and other services. Together with many HP customers, we are exploring novel ways to use Big Walls to address issues in areas such as disaster recovery, emergency response, hospital wards, city monitoring and management, mergers and acquisition planning, product planning, and supply chain management. We are creating a next generation capability by using Big Walls together with mobile devices, touch surfaces, 3D data visualization techniques, and techniques for displaying, manipulating and visualizing large amounts of unstructured data. In addition to our research, we participate in the real world by providing support for HP’s worldwide operations centers and the HP supported gallery at the Newseum due to open in 2012.

Links to entertainment reviews
HP 3D Live: CES Earth Wind and Fire concert

http://blogs.forbes.com/oliverchiang/2011/01/08/ces-hp-believes-in-3-d-too-but-on-a-larger-scale-much-larger/

http://www.monstercable.com/events/ces2011/3dexperience.asp

http://h20435.www2.hp.com/t5/The-Next-Bench-Blog/Earth-Wind-and-Fire-Going-Large-Live-and-in-3D-at-CES/ba-p/60745

http://www.hardwaregeeks.com/index.php/site/comments/hp_streams_concert_live_in_3d/

http://venturebeat.com/2011/01/14/hp-streams-earth-wind-fire-in-live-3d-on-a-huge-screen/

2010 Sundance Film Festival

http://philmckinney.com/archives/2010/01/sundance-and-3d.html

http://h20435.www2.hp.com/t5/The-Next-Bench-Blog/Stories-in-3D/ba-p/52743

HP Newseum announcements

http://www.hp.com/hpinfo/newsroom/press/2010/100930a.html

http://www.newseum.org/news/2010/09/hp-announcement.html

My primary research interest is in lifelong machine learning, including:
- statistical methods for knowledge transfer,
- methods for scaling transfer learning to long sequences of tasks,
- sparse representations for transfer learning,
- self-directed goal-driven learning,
- interactive learning in collaboration with a user, and
- the application of these methods to problems in medicine, search & rescue, and space exploration.
I am also conducting research on community detection in relational networks, and AI education.

My research focuses on the development of mathematical models and algorithms for estimating evolutionary history in Biology and Historical Linguistics. The main objective is to develop methods that produce much more accurate estimations of evolutionary history than can be obtained using existing tools. Our group is distinguished from many other groups in computational biology due to our focus on ultra-large datasets, with up to 500,000 sequences. We use real data and perform simulations to evaluate the performance of methods that we develop. This research area inolves mathematics, probability, statistics, computer science, and intensive collaborations with domain specialists. My current research is funded by two grants from the National Science Foundation, one an Assembling the Tree of Life (ATOL) grant for simultaneous estimation of alignments and trees, and another for estimating species trees from gene trees. I also have an active interest in metagenomic analysis. No background in Biology is needed. However, good mathematical intuition, and skill in algorithm and software development are important.

Current research interests include Bayesian and Bayesian nonparametric methods for:
- modeling nonlinear dynamical phenomena
- identifying latent group structure between multiple time series
- scaling to highly multivariate time series by inferring conditional independencies or harnessing lower dimensional embeddings of dynamics
- exploiting multiple sources and types of data
- efficient inference (optimization-based methods, data-driven MCMC, SMC methods for online inference)

I am interested in shape / data analysis using geometric and topological methods, with applications ranging from graphics, visualization, bio-informatics, and manifold learning. In particular, my research aims to integrate geometric and topological ideas to develop both theoretical foundations and practical algorithms for information estimation, feature identification and abstraction, shape characterization, matching, and shape deformation. I am also interested in exploring how structures and results developed in computational geometry and topology can be used and extended to complement / augment statistical data analysis methods.

I work mostly in two areas Computational Geometry and Computational Biology.

I have broad HCI and CSCW interests. I am interested in cognitive/social design principles underlying the design of systems in the following areas:
- large scale social and collaborative systems (e.g. communities)
- personal information management
- digital support for memory/lifelogging
- interactive information retrieval

My university is just 5 mins from the Pacific in a beautiful part of n. California

With the rapid growth of mobile devices, sensing techniques under mobile conditions become more and more important. I am interested in sensor signal processing, analysis, and fusion for portable and real-time health care monitoring and clinical measurements. I am also interested in techniques for smart 3D sensing, interaction, and display. Other directions include security and biometrics, signal retargeting, enhancement, compression, and retrieval, driving safety monitoring, and smart and flexible robot training.

Over the last three decades Prof. Crutchfield has worked in the areas of nonlinear dynamics, solid-state physics, astrophysics, fluid mechanics, critical phenomena and phase transitions, chaos, and pattern formation. His current research interests center on computational mechanics, the physics of complexity, statistical inference for nonlinear processes, genetic algorithms, evolutionary theory, machine learning, quantum dynamics, and distributed intelligence. He has published over 130 papers in these areas, including the following recent, related publications. Most are available from his website: http://csc.ucdavis.edu/~chaos/.

Within the fields of neuroscience and machine perception and learning, I work on ways to discover structure automatically in complex sensory environments, and on relationships between neurobiology, signal processing, and information and control theory. My early work in this field pioneered in the use of information theory to devise novel principles for neural network learning and for understanding biological sensory processing.

My current interests include the application of “deep learning” neural networks (e.g., multi-stage Restricted Boltzmann Machines and related algorithms) to detecting rare anomalous behavior in large financial datasets. A prospective mentee should have a PhD with substantial knowledge of and practical experience with neural networks. Some experience with deep learning NNs is especially desirable.

My main theoretical focus is on the potential connections between deep learning NNs and the neural organization of cerebral cortex. I seek principles that may account for aspects of known biological neural architecture, and insights from biology that may enhance the power of artificial NNs.

My research interests cover a wide variety of topics including Smarter Web, Web Task Modeling, Web Mining and Web accessibility. The distinguishing aspect of my research is that I seek efficient and theoretically sound techniques to solve a problem in a principled way and validate the techniques by implementing them in real systems. One of the key features of my research is to improve the usability of a Web based system by identifying and presenting relevant information to the user. I have developed information presentation techniques (e.g., contextual browsing, model-directed navigation, conversational interface) to improve usability of such a system. I applied those techniques to reduce information overload problem in different areas, such as Web Access using different modalities, Web automation and testing, information retrieval, Email and collaborative task management, social media analysis. Another important aspect of my research is that I have applied machine learning algorithms to get efficient solutions in many instances, e.g., learning relevance model of a Web user, Web task models, Web accessibility models, learning models for website testing.

My research group creates new software and algorithms which enable robots to effectively assist physicians and automatically accomplish healthcare tasks. We focus on developing motion planning algorithms and physically-based simulations with applications to robot-assisted surgery, treatment planning, medical image registration, personal assistance, and physician training. Our research spans the following areas:

(*) Motion Planning for Healthcare Robotics: Our objective is to compute actions that enable a robot to automatically accomplish clinical or assistive tasks. Using sampling-based and geometric approaches, we develop algorithms to automatically maneuver medical devices such as robotic surgical assistants or steerable needles around anatomical obstacles to difficult to reach sites, enabling new surgical and interventional procedures. Our research identifies and exploits parallels between medical problems and traditional robot motion planning problems while explicitly considering the complexity and uncertainty inherent in healthcare applications.

(*) Physically-based Medical Simulation: Human soft tissues are heterogeneous and have nonlinear properties, resulting in complex deformations during clinical procedures. Using finite element methods and mesh maintenance algorithms, we are developing simulations of soft tissues and their interaction with medical devices. These simulations can assist physicians in registering diagnostic and treatment images obtained at different times, and can also be used for interactive physician training and procedure planning.

Research in my lab focuses on understanding the information processing strategies employed by the visual system for tasks such as object recognition and scene analysis. Computer scientists have long sought to emulate the abilities of the visual system in digital computers, but achieving performance anywhere close that exhibited by biological vision systems has proven elusive. Our approach is based on studying the response properties of neurons in the brain, and attempting to construct mathematical models that can describe what neurons are doing in terms of a functional theory of vision. The aim of this work is not only to advance our understanding of the brain, but also to devise new algorithms for image analysis and recognition based on how brains work.

Anything to do with information. Have been involved in > 110 research grants, giving hundreds of papers and presentations, initially on information retrieval, including computational linguistics and reasoning under uncertainty. Also worked on multimedia, educational technologies, digital repositories, and extending services related to electronic theses and dissertations (including analyzing references and classifying works and parts, see www.ndltd.org), with machine learning. One of the founders of the digital libraries area, developing formal foundations, prototype systems, user studies especially on interfaces and information visualization, and applications including to archaeology, physics, and education. Leading the Crisis, Tragedy, and Recovery network (www.ctrnet.net), including working with ontologies, focus groups, twitter, focused crawling, filtering.

We recognize visual objects with such ease that it is easy to overlook what an impressive computational feat this represents. Any given object in the world can cast an effectively infinite number of different images onto the retina, depending on its position relative to the viewer, the configuration of light sources, and the presence of other objects in the visual field. In spite of this extreme variation, biological visual systems are able to effortlessly recognize at least hundreds of thousands of distinct object classes—a feat that no current artificial system can come close to achieving.

My laboratory seeks to understand the computational underpinnings of object recognition through a concerted effort on two fronts. First, we endeavor to understand the workings of biological visual systems using a variety of experimental techniques, ranging from microelectrode recordings to visual psychophysics. Second, we attempt to instantiate what we have learned into artificial object recognition systems, leveraging recent advances in parallel computing to build systems that begin to approach the scale of natural systems. By combining reverse- and forward-engineering approaches, we hope to accelerate progress in both domains.

I’m applying my bioinformatics research to solve problems in systems biology with implications to medical conditions such as cancers and addictions. My interest include cancer genetic and neurogenetic discovery based on SNPs, gene expression microarrays, protemics data, and biomedical ontologies. Specifically, I utilize machine learning tools, information theory, and holistic statistical constructs such as Bayesian Networks. My research involves both leveraging well established methodologies and developing novel algorithms to push the limits of biomedical informatics.

Ginkgo BioWorks is a young company out of MIT with the mission of making biology easier to engineer. We engineer organisms to solve challenges across a range of industries from fuels to pharmaceutical production. We aren’t trying to study biology, we are trying to build it – constructing, editing, and redesigning the living world. Our bioengineers make use of an in-house pipeline of synthetic biology technologies to design and build new organisms.

We are looking for a computing postdoc who is passionate about developing CAD tools for the engineering of organisms. You probably have a background in comparative genomics, computational biology, metagenomics, or other similar areas. You have likely been building computational tools to study natural biology — but at Ginkgo you would have the opportunity to apply your skills and tools to the challenge of engineering new organisms.

At Ginkgo, our organism engineers have immediate demands for expanded tools to support organism design. We believe that we provide a unique environment for someone to cut their teeth on real CAD problems in biological engineering. Lastly, we’d be happy to serve as mentors for CIFellows if you are pursuing that fellowship.

Our group at MIT aims to further our understanding of the human genome by computational integration of large-scale functional and comparative genomics datasets. We develop algorithmic, statistical, and machine learning methods to interpret the functional elements encoded in the human genome, reconstruct the regulatory circuits they define, and understand their evolutionary mechanisms.

(1) We use comparative genomics of multiple related species to recognize evolutionary signatures of protein-coding genes, RNA structures, microRNAs, regulatory motifs, and individual regulatory elements.

(2) We use combinations of epigenetic modifications to define chromatin states associated with distinct functions, including promoter, enhancer, transcribed, and repressed regions, each with distinct functional properties.

(3) We use dynamics of functional elements across many cell types to link regulatory regions to their target genes, predict activators and repressors, and cell type specific regulatory action.

(4) We combine these evolutionary, chromatin, and activity signatures to dramatically expand the annotation of the non-coding genome, elucidate the regulatory circuitry of the human and fly genomes, and to revisit previously uncharacterized disease-associated variants, providing mechanistic insights into their likely molecular roles.

I do work research in social computing and social media. My approach to research typically centers around using computational methods (e.g., applied NLP) to both understand the social internet and build new systems for it.

Current topics:

* Building new email interactions
* Building new interaction techniques for social networking sites
* Analyzing diffusion through networks over long spans of time
* Analyzing online social behavior through theoretical lenses

My work i is broadly at the intersection of HCI and Information Retrieval/Data Mining. I generally work with extremely large datasets of text (Web, large collections of scientific texts, historical collections of books), behavior (click and search logs), and networks (social systems, Web, P2P, etc.). My work tends to have both an empirical descriptive/modeling component as well as a system-building component (leveraging what know to build better systems). Current areas of focus include:

- Temporal-informatics – providing front ends to users who have data-extraction needs for time-varying data (e.g., http://www.cond.org/zoetrope.html)

- Meme mutation – how information mutates as it moves through networks

- Scientific data evolution – mining large collections of scientific literature to detect evolution and emergence

- Collaborative analytics/visualization – building systems to support users without a great deal of statistical and visual literacy in understanding data

- Privacy – how perceptions shape behavior

Speech recognition, spoken language translation, situated spoken language understanding, machine learning

My work lies broadly at the intersection of HCI, interactive IR and cognitive psychology. The overarching goal is to use the implicit user data in personalization of interaction with information systems.

Current areas of focus include application of cognitive psychology and neuroscience to information seeking (IR and web navigation). The approach is to use rich user interaction logs, eye-tracking and psycho-physiological signals (such as EEG, GSR) to infer user’s states (e.g., cognitive load, affective states) and user’s reaction to information (e.g., decisions).

There are opportunities to :
* study information seeking behavior in novel ways;
* work on creating novel search and navigation interfaces;
* work with large psycho-physiological data sets and apply signal processing and machine learning techniques.

Worth noting: My university is just about an hour away from New York City.

My current research interests lie in computer vision and machine learning. In computer vision, I am interested in facial processing and analysis, object tracking, event/activity recognition, and image segmentation. In machine learning, I am specifically interested in theoretical developments in probabilistic graphical models (e.g. Bayesian Networks and MRF, HMM) including learning, inference, and their applications in different fields, in particular in computer vision. I am also interested in decision making and information fusion based on graphical models such as the influence diagrams. Refer to my website for details

The following describes the projects I am currently most interested in.

The field of machine learning increasingly leverages graph representations and graph-based algorithms in a variety of applications and problem settings. For instance, semi-supervised learning, dimensionality reduction, and unsupervised clustering problems are frequently solved by casting data points as nodes in a graph and then applying a graph-theoretic algorithm. Similarly, probabilistic inference problems are solved by casting random variables as nodes in a graph and then applying graph algorithms such as message passing, tree approximations and other variants. Most of these learning and inference graph problems are NP-hard in general. Similarly, various combinatorial graph problems such as graph coloring, maximum stable set and maximum clique are NP-hard in general. However, in combinatorics, scientists have identified a large family of graphs known as perfect graphs where combinatorial problems become tractable and admit polynomial time algorithms. We hope to investiga te a variety of hard learning problems and, by compiling them into combinatorial problems on perfect graphs, identify situations in which they may be solved exactly or approximated well. We plan to investigate various algorithms for exact solution including linear programming, semidefinite programming and message passing.

The first type of learning setting involves inference in graphical models: graph-based representations for encoding a set of conditional independence properties that hold in a probability distribution p(x_1,\ldots,x_n) defined over a set of random variables. Two important canonical computational problems therein are MAP Estimation (finding the configuration that maximizes p(x_1,\ldots,x_n)) and Marginal Inference (finding the marginal configuration of variables, such as p(x_i)). Both of these are NP-hard in general. However, it is possible to compile the graphical model into another graph and solve an equivalent maximum stable set problem. If this graph happens to be a perfect graph, the inference problems are tractable. We hope to follow this general approach to identify sufficient conditions for tractable graphical model inference.

The second type of learning setting to investigate involves operations on data graphs: graph-based representations which encode pairwise similarity relationships between pairs of objects {x_1,\ldots, x_n}. Two canonical computational problems therein are Clustering (finding multiple disjoint subsets of the data with relatively high self-similarity) and Anomaly Detection (finding a single subset of data with relatively high self-similarity).

We aim to develop methods of solving these problems by ensuring that the data graph constructed from these samples is perfect. This allows the clustering problem to be mapped to a graph coloring problem and allows the anomaly detection problem to be mapped to a maximum clique problem. Both of these problems are solvable in polynomial time for perfect graphs using convex programming methods.

Please apply for a CI Fellow position with me and if you are in the New York area, please try to drop in for a visit.

The goal of the Galois project is to develop programming notations, compilers and runtime systems for programming multicore processors. We focus on irregular algorithms, which are algorithms in which the key data structures are sparse graphs, trees, sets, etc. This class of algorithms is very general, and includes mesh generation, refinement and partitioning algorithms, SAT solvers, machine learning algorithms, n-body methods etc. Our work has appeared recently in top conferences like PLDI 2011, POPL 2011, OOPSLA 2010, and ASPLOS 2011. Our group has six PhD students and three post-docs, and we are always looking for new talent!

My research interests are in reinforcement learning, structured prediction, transfer learning, and machine learning for planning and natural language processing. I am interested in a wide range of aspects of these problems from fundamental theoretical aspects to real world applications in real-time strategy games, emergency response, curriculum scheduling, and medical informatics.

My research primarily focuses on enabling synergetic autonomy in human-robot interactions in dynamic real-world application domains such as elderly care and surveillance. Each robot operates autonomously as far as possible, using algorithms to: learn environmental models based on sensory inputs; adapt the learned models in response to environmental changes; tailor sensing and information processing to the task at hand; and robustly merge the information obtained from teammates. At the same time, the robot acquires and fully utilizes limited high-level information from humans through a variety of interaction modalities, e.g., vision, natural language and touch-screen visual displays.

I am also interested in designing and adapting Bayesian reasoning algorithms to address real-world challenges characterized by a significant amount of uncertainty. Some (current) examples of applications are: medical data processing, HRI/HCI interface design, and adaptive learning technologies for training teachers and promoting learning among students with disabilities.

My group works in computer vision with an emphasis on the interpretation of video and the understanding of human behavior, in particular the social behavior of children. I am leading a large multi-institution effort to develop the science and technology of behavior imaging— methods for sensing and modeling behavior to support the diagnosis, treatment, and understanding of developmental disorders such as autism. A related area of interest is video interpretation and dynamic scene understanding, where we have developed novel video segmentation methods based on temporal causal analysis. A third area is perception for robotics, with specific projects in autonomous high-speed driving (the autonomous rally car project) and manipulation in clutter using the PR-2 robot from Willow Garage. Projects are available in all of these areas.

I am interested in how we can get robots and people to work more effectively together. Can we build control interfaces to robots that allow non-specialists to more effectively control them and get them to do useful work? Can we add social gestures to mobile robots and make them more efficient as they go about their tasks around humans? Can we decode signals from the human brain, and use these to control robots? In all of these areas, I am interested in the use of shared autonomy control, where the robots are as autonomous as possible, but are able to ask for help when this autonomy fails.

My research is in the areas of computer vision, image processing, and pattern recognition. I develop robust techniques and algorithms for image feature extraction and classification, and the application of these techniques to different real world problems. Representative projects I have worked on include document image watermarking and authentication, video scene analysis and classification, fingerprint image verification, license plate recognition, medical cell image analysis, 3-D object recognition, infrared target classification, road crack image analysis, and character recognition.

Raymond J. Mooney is a Professor in the Department of Computer Science at the University of Texas at Austin. He received his Ph.D. in 1988 from the University of Illinois at Urbana/Champaign. He is an author of over 150 published research papers, primarily in the areas of machine learning and natural language processing. He is the current President of the International Machine Learning Society, was program co-chair for the 2006 AAAI Conference on Artificial Intelligence, general chair of the 2005 Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing, and co-chair of the 1990 International Conference on Machine Learning. He is a Fellow of the Association for Computing Machinery and the American Association for Artificial Intelligence and recipient of best paper awards from the National Conference on Artificial Intelligence, the SIGKDD International Conference on Knowledge Discovery and Data Mining, the International Conference on Machine Learning, and the Annual Meeting of the Association for Computational Linguistics. His recent research has focused on learning for natural-language processing, connecting language and perception, statistical relational learning, abductive reasoning, transfer learning, and active learning.

I am broadly interested in topics at the intersection of HCI and AI and Machine Learning. Major areas of focus include:

– Systems that adapt the interaction to the abilities (motor, perceptual, cognitive) of their users; this is to accommodate differences due to activity, health condition, context, culture, and bias;

– Interactive machine learning: we are developing novel interactions and algorithms that enable machines to learn from a small number of interactions with the user;

– Creativity support tools: how can personalized software help people explore the design space more efficiently, create more diverse alternatives, and improve self-efficacy?

Design, modeling, measurement, and analysis of conventional and biologically-inspired high-performance computer architectures and their interaction with state-of-the-art optimizing compilation systems, modern operating systems, and scientific, commercial, and intelligent applications.

Large complex systems require real time operations management. Some example include domains such as managing transportation networks, managing large computer systems and massive data centers, managing electric grids, etc.

Since the underlying mathematical models for these complex systems are intractable, real time management of such systems require us to build systems that can process large amount of data and take decisions in real time. The system will need to be able to efficiently process stored and streaming data, structured and unstructured data, and event and sensor data.

Given the problem above, we are interested in building new algorithms and techniques in areas such as complex event processing, data mining and machine learning, information management. Besides the algorithmic challenge we are also interested in building, designing and optimizing a large scale system that can implement these algorithms and techniques for various industry verticals such as healthcare, education, energy, etc.

Our focus is the development of theory and tools for processing of signals that arise in today’s growing array of different applications—and pose challenges for the traditional signal processing techniques. Our focus is the development of data-driven methods for the analysis and fusion of medical imaging data. Our collaborations include Johns Hopkins Medical Institutions and the Mind Research Network.

The Interface Ecology Lab imagines and develops integral and sensitive human-centered computing to support nuanced and exciting aspects of life, including how we form and express ideas and intentions, how we learn and innovate, how we gesture and communicate, how we coordinate and cooperate, how we participate and share, how we function under stress, and how we respond to crises. We connect diverse methodologies, engaging an interface ecosystems approach to engage computing in holistic avenues of human experience.

We have just developed an innovative multi-finger sensor, ZeroTouch. We are now poised to integrate ZeroTouch into interactive environments, developing new embodied bi-manual and free air interaction techniques and experiences.

We have released the open source meta-metadata language and architecture, for authoring platform-independent wrappers for heterogeneous information sources. Meta-metadata supports information extraction and knowledge integration. It facilitates writing programs that enable new human experiences of information visualization, physically-based modeling, and embodied interaction. Research here involves new knowledge semantics and modeling, on the one hand, and new interactive visual applications, on the other. Social media and digital libraries are connected.

We are developing special relationships with disaster response organizations. Texas Task Force 1 engages in major disasters, such as 9/11, and hurricanes Katrina, Rita, and Ike. TEEX Disaster Preparedness and Response runs an internationally-renowned academy for first responders. We are engaged both in developing zero fidelity simulation games for education, and in creating innovative multi-surface information systems for crisis response.

The typical search interface is great for finding a single element of information, but weak for information-based ideation tasks, in which the human goal is to to develop ideas with support and stimulus from information. The combinFormation platform reconceptualizes information-based ideation support as mixed-initiative information composition, integrating browse-search-collect-visualize-and-organize, while representing each collection as a connected whole. To validate composition, information-based ideation metrics and grounded theory are extended and synthesized, developing formative and summative evaluation methodologies. Application contexts include thesis writing in computer science, and design in architecture and mechanical engineering.

My laboratory is interested in key challenges in Systems Biology and Medicine. Our larger objective is to decipher biological mechanisms, reconstruct networks, predict phenotypes and build quantitative systems models. Towards this end we continue to develop computational and some experimental methods for integrative analysis of multiple types of biological data and for reconstruction of biochemical networks. It is increasingly become clear that time series data in cellular biology and longitudinal data in mammalian in vivo biology are essential for understanding complex phenotypes and we are developing methods to analyze these data. And most importantly, I subscribe to the view that context-free biology is content-free biology!

I am interested in advancing the state of the art in information retrieval and natural language processing at Web scale. I am also interested in using large-scale repositories of collaboratively-generated content (e.g., all of Wikipedia) for improving text processing and artificial intelligence applications.

For additional details, please see my Web site.

Computational Imaging, Computational Photography and Display, Computer Vision, Medical Imaging, 3D imaging, Image,Video and Light-Field Processing

I work to understand motion and change in video — separating changes in seasons from changes over a day in outdoor time-lapse, capturing the motion patterns in an MRI video of a heart, or parsing the surveillance video of a intersection into traffic cycles. I mostly work on data-driven methods that could scale to work with (for example) all the webcams connected to the web, or within large scale, image based Citizen Science applications.

I research in artificial intelligence and in robotics. I have a long-term research goal in intelligent (multiple) robots and agents that are part of human environments. I am interested in algorithms for planning under uncertainty, learning, and execution also for multiagent and multi-robot systems.

I am currently particularly interested in the integration of physical, web-based, and human-provided information for service robots to effectively function in indoor environments.

I am working on exciting computational biology projects related to synthetic biology, designing algorithms and tools that enable life scientists to create synthetic genomes. Working with virologists, we have already designed and synthesized virus genome sequences to serve as vaccines. Our work was reported in Science and Nature Biotechnology. We expect the next generation of tools to help design artificial chromosomes and cells performing unique functions, using existing genes, transcription factors and pathways as optimized design components.

I am involved in a number of other bioinformatics projects, including genomic sequence classification using machine learning, repeat mapping from sparse data, and algorithms for optimizing protein encodings adhering to a variety of constraints, including secondary structure, codon bias and other factors. I am also exploring the area of copy number variation detection from fragmented genomic sequence data, as well as using grid computing for time/space consuming applications, such as short read genome assembly.

I also have an active collaboration in the field of video streaming and video pattern matching. We are exploring efficient algorithms for simultaneously streaming compressed video through fixed bandwidth channels. We are also working in identifying patterns in the information content of video streams.

There are several potential research projects within the linqs group at UMD (www.cs.umd.edu/linqs) ranging from core machine learning and probabilistic modeling, to applications in social network, personalized medicine and information alignment. We have a large research group and a CI Fellow would gain experience in mentoring undergraduate and graduate students, grant writing, and teaching (if interested).

Seeking postdoctoral fellow to work on predictive dynamic models of language and nonverbal behavior and integrating them into Embodied Conversational Agents. The candidate should have a background in Computational Linguistics (NLG, Dialogue); should be willing to do significant reading into current approaches (both symbolic and probabalistic) to generating dialogue, should be capable of incorporating recent linguistic and psychological insights into the dyadic, dynamic nature of dialogue, and the integral role played by eye gaze, gesture, posture; and should be willing and able, therefore, to develop more innovative and adequate approaches to generating dialogue. The result will be a radically new way of generating embodied dialogue moves, for an embodied agent that can be more tightly integrated into collaborative scenarios with humans. Applications will include embodied peers that teach literacy and scientific reasoning behaviors, embodied peers that scaffold social behavior in children with autism, and embodied agents that give directions in real and virtual environments.

Professor Cassell is currently faculty at Northwestern University, but will be joining the Carnegie Mellon Human-Computer Interaction Institute in Fall, and so the postdoctoral candidate would be located at CMU HCII in Pittsburgh, with access to colleagues in Natural Language Technologies, Machine Learning, and Graphics.

Please see my web page.

We study all aspects of Computational molecular biology as it relates to proteins, from sequence to structure to function. We have particular interests in remote homology detection and in protein-protein interaction networks. The toolbox we need to solve these problems comes from algorithms, machine learning and even graphics/ visualization. I am happy to supervise anyone with a strong background in any of these three areas even if prior experience with biological problem domains is limited. Many opportunities to learn about this area and develop new interdisciplinary collaborations.

My main area of research is Natural Language Processing (NLP), and its application to human-computer interaction, educational technology, and multimedia systems. My goal is to use NLP to support both education and instruction, and collaboration between human or artificial agents. My group focuses specifically on semantics, and discourse and dialogue processing. All my research has its empirical foundations in both qualitative and quantitative corpus analysis, including data mining techniques. In the last decade I have worked on language interfaces for educational technology, on discourse structure / dialogue act recognition, and on summarization. I am currently exploring a new exciting application area, that of health care, e.g. for assistive robotics, decision support systems for patients, and medical education.

I recommend you go to the Healthcare Robotics Lab website (http://healthcare-robotics.com) to learn more about my lab’s research. There are many opportunities for postdoctoral researchers to make high-impact contributions. If you have any questions, please do not hesitate to contact me.

My lab’s research seeks to advance the capabilities of real robots so that they can provide valued assistance to people in the context of healthcare. Specifically, we work with semi-autonomous mobile robots that physically manipulate the world (mobile manipulators). Our current efforts include research into human-robot interaction, autonomous mobile manipulation, machine perception, machine learning, and haptics.

The following three large-scale efforts are especially exciting to us and we would welcome a postdoc to become involved:

1) Assistive Mobile Manipulation for Older Adults at Home: This is the PR2 Beta Program project at Georgia Tech for which I’m the principal investigator. It involves several initiatives with the overall goal of providing assistance to older adults in their homes. Our PR2 is currently in the Aware Home, where it will be interacting with older adults this summer (2011).

2) We have a new project that is not yet public that is very exciting in terms of helping people with severe motor impairments. Please contact me for details.

3) Haptically-guided Manipulation within Extreme Clutter: We are investigating the possibility of a new foundation for robot manipulation using compliant actuation and whole-body skin. This project is also new and not much is public, so please contact me for details.

My lab is multi-disciplinary with members from computer science, biomedical engineering, electrical engineering, and mechanical engineering. We are also collaborative. We are active in the Robotics PhD program and the Center for Robotics and Intelligent Machines (RIM@GT), and we work closely with other labs on campus, including the Human Factors and Aging Lab in the School of Psychology and the Socially Intelligent Machines Lab in the College of Computing. Although much of our research has a long time horizon, we actively engage industry and look for opportunities to make an impact.

We have several state-of-the-art robots, including a PR2 from Willow Garage, the custom humanoid robot Cody, and a variety of smaller assistive robots that we have developed. By the end of the summer (2011), we also expect to have an additional sophisticated human-scale mobile manipulator. Our other facilities include rapid prototyping equipment (3D printer and laser cutter), a machine shop, and a force and motion capture studio.

Best of luck with your application.

- Charlie

I am interested in working with fellows in two distinct areas of research. The first is machine learning as it applied to both structured (relational) data and to time series data. The second statistical relational models and structured inference algorithms for them (so called lifted inference), especially but not restricted to models  rising from stochastic planning problems. Please see web site for examples of recent work.

Two sample projects involve
1. The development of methods for the inference of the identity of people observed through a network of cameras and possibly real-time, low-cost range sensors.
2. The analysis of video sequences over many frames.
These projects are briefly described below.

1. Cameras and sensors are deployed in a large building or other facility, and record video of people walking by. Observations are compared pairwise through metrics of similarity of appearance, and probabilistic models capture the likelihood that the same individual is observed at two given locations at times separated by a given interval. A directed graph connects two observations A and B if (i) A occurs before B, and (ii) it is possible to establish the probability p_AB that A and B correspond to the same person. Determining identity then amounts to partitioning the graph into maximal sets of observations pertaining to different people.
Open research questions involve both image and video analysis to compute the probabilities p_AB, and inference and graph partitioning methods to determine identities.

2. The images in a video sequence, stacked together, form a parallelepiped with dimensions corresponding to image columns (x),
image rows (y), and frame number (t). For brevity, we call this volume
a “video cube.” Every point (x, y, t) in the cube is ideally connected to the corresponding point (x’, y’, t+1) in the subsequent frame. These connections form a vector field (u, v, 1) that is called the “motion field.” The first goal of this project is to find occlusion boundaries, that is, surfaces within the video cube across which the motion field is discontinuous. The second goal is to find dense approximations to the integral curves of the motion field. These are formed by following the motion field vectors throughout the cube.
In contrast with a large portion of previous work, we are interested in methods that analyze long sequences, and our focus is on understanding the geometry of occlusion boundaries as surfaces in (x, y, t).

The postdoc would join a group where these and other computer vision projects are in progress, and would have ample flexibility in determining his or her own research agenda, in collaboration with my graduate students and me. We have weekly individual meetings, a weekly reading group, and a weekly group meeting with students and faculty in computer vision, artificial intelligence, and computational economics.

I will consider candidates in any of my research keyword areas.

Projects of current special interest in my group include:

* active learning to choose drug candidates during drug discovery
* active learning to fit scientific models, especially in astrophysics
* methods for finding groups or pattern of anomalies
* planning/navigation for active data collection in an anomaly finding task
* matrix completion methods in demand forecasting and collaborative filtering
* methods of smart regularization in learning
* learning dynamic models from scientific data where time/trajectory information is missing
* finding events in time series data from sensor networks, in particular for the CTBTO

The research goal of the Music Computation and Cognition group at USC is to systematically study, computationally model, and scientifically explain human abilities in music perception and cognition, and in music making, such as music performance, improvisation, and composition.

Research projects at the laboratory address the three main areas of music analysis, performance, and composition/improvisation. They include:

• decoding and generation of decisions and gestures in music performance;
• analysis of musical ensemble in networked music performances;
• multi-modal interaction in human-machine musical improvisation; and,
• automatic analysis and visualization of musical structure.

The MuCoaCo website — http://www-bcf.usc.edu/~mucoaco — and the MuCoaCo blog — http://mucoaco.blogspot.com — contain detailed project descriptions, video demonstrations, and lists of publications and awards. The broad areas spanned by the research projects are mirrored in the seminar course on Topics in Engineering Approaches to Music Cognition — http://www-scf.usc.edu/~ise575 .

The successful postdoctoral researcher is expected to engage in music computing research, and to have completed graduate-level training in at least one of the following quantitative analysis areas: mathematics, computer science, operations research, signal processing, statistics, or neuroscience. It is highly advantageous for the candidate to possess graduate-equivalent practical or theoretical knowledge in music analysis, performance, composition, or music perception and cognition. Preference is given to individuals with prior experience in music computing research.

MuCoaCo affiliates have had backgrounds in Computer Science, Electrical Engineering, Industrial and Systems Engineering, Music Composition, Music Performance, Neuroscience, and Psychology.

I lead the Teamcore research group, which is focused on applying research in multiagent systems (MAS) to real-world problems and advancing fundamental research to address these needs. We are particularly interested in real-world problems of security, safety and sustainability. As such, we explore the full spectrum of approaches in MAS, making necessary breakthroughs in the one that best suits the application at hand.

Key examples of this use-inspired research include our ARMOR, IRIS and GUARDS systems. While ARMOR is focused on a game-theoretic application for security scheduling at Los Angeles International Airport (LAX), IRIS does game-theoretic scheduling for the FAMS (Federal Air Marshals service) and GUARDS for the TSA (Transportation Security Administration). In this work, we created new algorithms for solving a class of game-theoretic games known as Stackelberg games to address efficiency concerns with existing techniques. Only in overcoming these challenges could we successfully produce a practical application for real-world use. Today, the ARMOR system has received numerous accolades and has been in use at LAX since August 2007. IRIS has been in use since October 2009 and GUARDS is being evaluated by the TSA for national deployment.

Our research in multiagent systems includes work in Belief-Desire-Intentions (BDI) systems, in Distributed Constraint Reasoning (DCR), and in Decision Theoretic and Game Theoretic approaches, a key aspect of our research is the interedisciplinary partnerships beyond Computer Science, such as with the Game theory and human behavior group.

Dr. Ungar’s research group develops scalable machine learning and text mining methods, including clustering, feature selection, and semisupervised and multitask learning for large bioinformatic and web-based problems. Example applications include semi-supervised methods for information extraction, multi-view learning for gene expression and fMRI data, and use of document and link structure for informing feature selection or transfer of knowledge between tasks.

Students work closely with co-mentors in the Schools of Medicine or of Management (Wharton).

Projects available:
Spectral methods for capturing word meaning and
Modeling the dynamics of scientific micro-communities

My work centers around the study of machine learning and inference methods to facilitate natural language understanding.

We are pursuing both fundamental questions in learning and inference and how they interact — most recently: Structure Learning, Constraints Driven Learning, Integer Linear Programming formulations for NLP, Constrained Conditional Models — and a range of natural language processing (NLP) problems — moving in the direction of semantics and natural language understanding, as well as ESL and Psycholinguistically motivated language acquisition.

My research interests are in data management, cloud computing, computer games, data mining, and data privacy.

The Linguistic and Assistive Technologies Laboratory (LATLab) at The City University of New York (CUNY), Queens College, conducts research in computational linguistics and human-computer interaction with a primary focus on accessibility applications and assistive technology for people with disabilities. In particular, we study the design of computer technology to address language and literacy impairments of people with disabilities – with a focus on two groups: people who have intellectual disabilities (ID) and people who are deaf.

Animations of American Sign Language for People who are Deaf:

Our research at CUNY has focused on sign language animation technologies for people who are deaf. American Sign Language (ASL) is a primary means of communication for one-half million people in the U.S, and it is a distinct language from English – with its own unique word-order, grammar, and vocabulary. A majority of deaf high school graduates in the U.S. have only a fourth-grade English reading level or below, yet many of these adults have sophisticated fluency in ASL. Therefore, software that can present information in the form of ASL animations would improve these individuals’ access to websites, communication, and information. Using our motion-capture recording studio, we are collecting a corpus of ASL performances, linguistically annotating them (with the help of a team of native ASL signers), and analyzing this data to create models to underlie ASL animation technologies (to make them more natural-moving and understandable for deaf viewers). We conduct experimental studies at the laboratory on a regular basis in which native ASL signers evaluate animations synthesized using alternative models. We have studies models of speed/timing, use of signing space, verb inflection, and we are beginning a project on facial expression in ASL. We are also interested in issues relating to linguistic generation of ASL and English-to-ASL machine translation. Further, we are interested in studying other forms of signing communication, including various forms of Signed English used by people who are deaf or hard-of-hearing in the U.S.

Text Readability Detection / Text Simplification for People with Intellectual Disabilities:

Many people with ID have limited English literacy and could benefit from software that could help them automatically identify information sources or websites that are at an appropriate level of difficulty or that could automatically simplify complex texts. Based on linguistic features of a text that can be automatically calculated through NLP software (the part-of-speech of different words, the syntactic parse-tree of the sentences, etc.), we are designing software to assign a difficulty score to a text to indicate whether it would be accessible for an adult with ID to read. We are also experimenting with various experimental designs to gather ground-truth data about how difficult-to-read texts are for these users; this is non-trivial because it can be difficult for these users to participate in a traditional comprehension experiments.

Other Projects:

We are interested in other projects related to applications of computational linguistics or human computer interaction to accessibility or assistive technology for people with disabilities.

Videogaming for Rehabilitation:
Striatal dopamine release during video game play may facilitate brain plasticity following perceptual learning. By combining visual distortion with video game environments and multimodal sensory feedback, we may be able to effect faster and more widespread learning during motor rehabilitation. Although videogaming techniques for rehabilitation have been applied to the upper limb with success, there are many disorders of the voice, speech, and swallowing system that may be amenable to this technique. Our work in this area is to develop and test novel videogame-based interventions for these disorders in order to improve the quality of life in individuals with sensorimotor disorders.

Novel Neurotechnology for Speech Assistance and Rehabilitation:
Rehabilitation of communication through novel human-machine-interfaces is the “next frontier” in neural technology. A multidisciplinary understanding of the neural dynamics during speech production and real-time signal processing techniques is essential for the advancement of these technologies. The long-term research agenda of our lab is to bridge speech science with engineering to design new approaches to advance speech human-machine-interfaces to a reliable and intuitive state for populations that are currently severely restricted in their ability to communicate.

* Generic Stream Data Mining: mining, integrating and tracking pattern evolution in data that arrives in huge quantities under extremely stringent memory constraints, making it necessary to process the data in only one sequential direction. Examples: web clickstream data, network activity/log data, astronomical observations, sensor data, social and text data (tweets, blog posts, news)

More Specific Activity Examples:

* Web Stream Mining: mining web data in particular mining data from multiple stream sources, such as web usage data, text and social networking data within a streaming environment (massive evolving data that requires scalability and need for adaptation of the learning algorithms and outputs as well as tracking and characterizing evolution patterns)

* Text Mining in Social Data Streams: mining and visualizing the inherent topics and events and tracking their evolution in text data such as content of Web blogs, fora and social networks within a streaming framework

My primary research interests lie at the intersection of privacy, security, data analysis, and data management. I am very interested in the field of secure information sharing and its various applications; as also the application of secure computation technologies to business processes such as supply chain management and optimization. Our lab looks at different aspects of privacy, such as privacy in graph structured data, privacy in location based services, medical health record privacy etc. We are also interested in the use of data mining techniques to enhance security, such as in Role Engineering. As such, our lab looks at fundamental data mining problems such as boolean matrix factorization and the use of these for solving domain issues such as those in access control.

Please visit my webpage for more information.

In robotics we are interested in motion planning with emphasis on high-dimensional systems and kinodynamic planning, planning from high-level temporal goals, assembly planning, reasoning with sensing and control uncertainty, flexible object manipulation, physical modeling, probabilistic methods in robotics, the geometry of motion, and the use of new enabling technologies. We also develop educational modules and distribute software for teaching motion planning.

In computational structural biology and bioinformatics we develop tools on high-performance systems to model protein structure and function, understand biomolecular interactions and help analyze, in the long run, the molecular machinery of the cell. We integrate sequence information with three-dimensional structural information to analyze and represent molecular flexibility and motion. Our work in computer-assisted drug design has also led us to investigations in systems biology.

– Binocular, multiple-view and video-based 3D reconstruction
- 3D reconstruction of dynamic scenes
- 3D shape representation and object recognition
- Analysis of large-scale range data
- Perceptual organization
- Manifold learning

We’re developing general systems for scalable Bayesian inference. We’re exploring a mixture of sampling and point estimation strategies including Hamiltonian Monte Carlo and variational methods.

General purpose tools under development include a general posterior sampler (along the line of BUGS), multiple imputation for missing data, and post-stratification for prediction.

We’re particularly interested in multilevel regression and factor models, with applications to prediction problems in epidemiology, climate modeling, and social science. I’m also interested in large scale probabilistic RNA alignment, expression estimation and pathway modeling.

Other team members include Andrew Gelman, Ben Goodrich, Matt Hoffman, and Michael Malecki.

I am interested in a broad range of questions in machine learning, applications and connections to cognition and other areas of science.

I am particularly interested in understanding and modeling non-linear structure of data in high dimensions, its applications to algorithms design and in the theoretical analysis and limits of such algorithms.

Jim Martin, along with Martha Palmer, directs the Center for Computational Language and EducAtion Research (CLEAR) at UC Boulder. CLEAR is dedicated to advancing fundamental Human Language Technology. Our current projects include improved statistical machine translation through the use of shallow semantics, application of NLP and machine learning techniques to electronic medical records, and automatic analysis of social media during natural and political crises.

I’m interested in solving machine learning. Of late, this includes research into mechanisms for efficient exploration algorithms, large scale learning, oracle learning algorithms, with side interests in structured learning and deep learning.

Our goal is to develop methods that will allow us analyze and ultimately to modify the computations that cells have evolved to carry out in response to signals from their environments.

I am interested in the areas of machine learning, automated planning, activity recognition, and the intersections of these areas. I am particularly interested in problems that require learning and inference in the context of rich knowledge representations. Two areas that I aim to make progress in over the next decade include:

1) Understanding the interplay between planning/inference and learning, where learning is used to both improve the quality and efficiency of planning/inference, and

2) Holistic activity understanding in complex videos of team sports, with the goal of developing new representations and associated inference and learning algorithms.

Candy Sidner is a research professor in the Computer Science dept. at Worcester Polytechnic Institute. I am interested in multi-modal computational models of verbal and non-verbal interaction, especially dialogue and discourse, with social and collaborative models, both for embedded conversational (on screen) agents and for human-robot interaction. My current research involves creating models of interaction for agents, both intelligent virtual agents and humanoid robots, and includes modelling relationships between people and agents, when the agents are “always-on” in people’s personal environments. My work includes the study of phenomena in human-human interaction, both dialogue and gestures, developing computational models and implementing and testing those models with everyday people. This work is joint with Prof. Chuck Rich at WPI and Prof. Tim Bickmore at Northeastern University.

I am interested in theoretical modeling and experimental understanding of multi-instrument measurement systems that deal with multi-dimensional multi-variate data, as well as in automation and design of computer-assisted systems that deal with large volumes and computational intensive processing of heterogeneous data. The main goal of my research and development is to automate information processing of repetitive, laborious and tedious analysis tasks, to scale computations of automated tasks, and to build decision-making systems that operate in automated or semi-automated mode.

Post-doctoral students and graduate students interested in working with me would be focusing on research and development of new algorithms, especially those that automate visual inspections and understanding scenes from spectral imagery, and those that scale with large volumes. The research of algorithms is based on deep understanding and modeling of instruments generating the images, the phenomena represented by images, and the processes applied to raw images. The contributions of the joint work would be in building bridges from raw data to information and to knowledge, as well as in understanding computational and algorithmic challenges for automated data-centric operations.

My recent research interests focus on crowdsourcing and on mining user-generated content on the Internet. I am especially interested in combining human intelligence with machine learning, and in building systems that provide quality assurance in crowdsourcing applications.

The following are my current projects:
1. Formal verification of embedded firmware (NSF funded)
2. Formal foundations of execution models for exascale computing (DARPA funded)
3. Analytics and activity-based intelligence (DARPA funded)
4. New paradigms in machine learning and complex event processing with applications to oil industry and agriculture (funded by BP and the State of Louisiana)

Recently completed projects:
1. Program synthesis and service-based systems (ONR and NRL funded)
2. Reliable middleware (NRL funded)

My research group focuses on developing representations and algorithms for learning causal models of complex systems of interacting components. These can be human communities (e.g., scientific communities, corporations, and government agencies), online social networks (Facebook, LinkedIn, and Twitter), peer production systems (Wikipedia, Stack Overflow), and complex computer systems (data centers, multiagent systems). We are developing radically new methods that assist human investigators in understanding the deep causal structure of these systems.

Research in my group deals with various aspects of the Web and increasingly the social Web, or Web 2.0. This includes analyzing and modeling the diffusion of information in social media, and building better applications for Web search, recommendation, navigation, and annotation. We study crowdsourcing applications and their incentives, such as games and services with purpose, to leverage the wisdom of the crowds for social computing tasks. Finally we study collaborative and distributed approaches to make the social Web less susceptible to abuses such as spam and manipulation.

Crowdsourcing, human computation, and collective intelligence are having profound implications for computing, posing many tantalizing research questions. Not only does the topic provide fertile land for research, it is also a topic where the time frame of a CI Fellowship can permit results with the significance and recognition that can help propel a career forward.

As one example direction, we now face micro-labor markets such as Amazon Mechanical Turk that allow us to write algorithms whose “subroutines” are people performing tasks at pennies a shot. How do we program this kind of “social computer”? What are the crowdsourcing analogs to the programming languages and IDEs that we rely on in traditional programming? I’m particularly (but not solely) interested in the role that machine learning can play here. For example, what’s the right way to do classification learning for problems that no longer need to assume that obtaining human labels on data is excessively costly, where people are now instead cheap albeit unreliable or differently motivated? This area not only poses new questions but also offers answers of relevance not just to the computational but also the behavioral sciences.

Human language is complex. Machine learning approaches to natural language processing (NLP) attempt to tackle its complexity by balancing background knowledge with data. My primary research interests are in understanding this balancing act, developing models of linguistic phenomena that can encode richer forms of background knowledge, and designing algorithms that can perform inference efficiently. My primary goals over the next five to ten years are: (1) To develop integrated models of language, taking advantage of raw multilingual text, insights drawn from linguistic typology, and grounding of language in other modalities (such as genetic pathways and simulated environments); (2) To develop machine learning paradigms to enable this, by combining different forms of knowledge into a single model.

I have been working on using Friedman’s functional gradient boosting method to learn Statistical Relational models such as RDNs and MLNs. In this approach, the problem of learning SRL models is turned into a series of relational functional-approximation problems. I am also interested in learning dynamic SRL models and causal models. Recently, I have been working with researchers at Wake Forest School of Medicine. In the first project, we are looking MRI scans of patients to identify tumors. The key challenge in this work is to identify the tumors at different locations of the brain given images of varying resolutions. I am looking at developing relational models that can generalize across different image resolutions and different positions. I have also been exploring the problems of inverse reinforcement learning (IRL) and imitation learning (IL) in relational domains where there is no explicit specification of a reward function. In these cases, the assumption is that there is an expert who provides trajectories as training examples for the agent.

information integration, query-evaluation algorithms, temporal reasoning,
automata-theoretic algorithms, firmware validation, protocol synthesis, constraint satisfaction, discrete techniques in robotic motion planning

My primary research interest is in the study of large-scale complex networks (such as technological, informational, and social networks) in order to build predictive models of such systems. Problems of interest span from knowledge representation issues to network classification and clustering, to attribution of macro-level behavior to micro-level elements. For more information, visit http://eliassi.org/pubs.html.

My research interests include
a) Automatic speech recognition: I work on all apsects of automatic speech recognition and speech-based biometrics. I am also one of the project leads for the popular CMU Sphinx open-source suite of speech tools. I also work on noise-robustness, multi-channel speech processing, and estimation of paralinguistic and contextual cues in speech. I am also interested in retrieval of spoken data and retrieval with spoken queries.

b) Computer audition: My interest lies in helping a computer understand the world through sound. This includes aspects of sound classification, source separation, and various decomposition techniques, but also higher-level inferences such as context/location determination (e.g. have we heard this guy before, and “where was this recorded and what else was happening”).

c) Multimedia analysis: This relates to (b) above, and includes image analyses and determination of audio-video association, semantic associations, and retrieval.

d) Security and privacy: Speech is probably the most popular form of communication between humans. The majority of the *private* communication in this world has also probably done through speech. Yet, in the large literature on techniques for privacy and security speech remains largely unaddressed. In this project I work on developing privacy-preserving techniques for speech processing. My group’s work includes aspects of signal processing, secure function evaluation and cryptography.