Research Projects

The new era of large-scale experimental methods in molecular biology has transformed it into an information-based science, making bioinformatics an integral part of genomic research. The research focus of the Laboratory of Bioinformatics and Functional Genomics is the development of integrated computational and experimental technologies for the study of gene function and regulation in biological systems through analysis, modeling, and visualization of heterogeneous biological data. The is is a joint laboratory with the Department of Computer Science and the Lewis-Sigler Institute for Integrative Genomics.

This project investigates how to build an efficient, high-quality content-based similarity search engine for feature-rich (non-text) data, which has dominated the increasing volume of digital information. The research topics include sketch construction, indexing for similarity search, distance functions for different feature-rich data types, integration with attribute-based search tools, content-addressable and searchable storage system, and Memex systems. The current toolkit is used to construct search systems for four data types including audio recordings, digital photos, 3D shapes, and genomic micro-array data.

The CertiCoq project aims to build a proven-correct compiler for dependently-typed, functional languages, such as Gallina—the core language of the Coq proof assistant. A proved-correct compiler consists of a high-level functional specification, machine-verified proofs of important properties, such as safety and correctness, and a mechanism to transport those proofs to the generated machine code. The project exposes both engineering challenges and foundational questions about compilers for dependently-typed languages.

My group develops algorithms for a diverse set of problems in computational molecular biology. We are particularly interested in predicting specificity in protein interactions and uncovering how molecular interactions and functions vary across context, organisms and individuals. We leverage high-throughput biological datasets in order to develop data-driven algorithms for predicting protein interactions and specificity; for analyzing biological networks in order to uncover cellular organization, functioning, and pathways; for uncovering protein functions via sequences and structures; and for analyzing proteomics and sequencing data. An appreciation of protein structure guides much of our research.

The Seung Lab uses techniques from machine learning and social computing to extract brain structure from light and electron microscopic images.

Cryptocurrencies and blockchains

Enterprise and data-center networks

We are currently developing methods for performing epigenome-wide scans for association of methylation status with phenotypes of interest.

EyeWire is a game to map the brain from Seung Lab at MIT. Anyone can play and you need no scientific background. Over 130,000 people from 145 countries already do. Together we are mapping the 3D structure of neurons; advancing our quest to understand ourselves.

Fairness and ethics in computing

FCMA: Full Correlation Matrix Analysis of Human Brains

Scalable causal consistency for wide-area data replication

ImageNet is an image database organized according to the WordNet hierarchy (currently only the nouns), in which each node of the hierarchy is depicted by hundreds and thousands of images. Currently we have an average of over five hundred images per node. We hope ImageNet will become a useful resource for researchers, educators, students and all of you who share our passion for pictures.

Internet architecture

Internet policy

Numerous decision problems reveal noisy, corrupt or otherwise partial feedback on choices. The Bandit Convex Optimization framework is a general methodology for tackling such missing-feedback problems, including the famed multi-armed-bandit problem and its generalizations. Applications include online routing, rank aggregation, matrix completion and more. Our research addresses efficient algorithms for bandit convex optimization, both in the form of general methodologies for exploration, as well as new techniques from optimization, and variance exploitation.

Inferring and Enforcing Security Policies in Web Applications

The Liberty Computer Architecture Research Group exploits unique opportunities exposed by considering the interaction of compilers and architectures to increase performance, to improve reliability, to reduce cost, to lower power, and to shorten the time to market of microprocessor systems. This objective is accomplished by providing critical computer architecture and compiler research, expertise, and prototypes to the community.

For much of my professional life, designing algorithms had been my thing. Then, one day, I watched a majestic flock of geese fly over Carnegie Lake and it dawned upon me that it had been their thing, too. Having been at it for 100 million years, even longer than I had, naturally their algorithmic genius surpassed mine. Undaunted, I resolved to catch up. The premise of my current research is that interpreting biological or social self-organized systems as "natural algorithms" brings upon them a fresh, new perspective ripe for inquiry. I believe that only the algorithm has the expressive power to model complex self-adaptive systems at the right levels of abstraction. Algorithms are the differential equations of the 21st century. Beyond its trite catchiness, this line serves to remind us that mankind's grasp of PDEs vastly exceeds its mastery of algorithms. The first order of business, therefore, is to build new analytical tools for natural algorithms.

The Network Programming Initiative supports research on languages, algorithms, and tools for network programming, and facilitates closer interactions with partners in industry and government.