To obtain scalable Bayesian inference methods, we develop algorithms to create compact “summaries” of large quantities of data. We can then quickly run standard inference algorithms on these summaries without needing to look at the whole dataset.
Our goal is to understand the illumination of an environment. By disentangling the illumination effect from other intrinsic properties (e.g. geometry, texture, color), we can better understand how human perceive the world. It also has several applications such as single image relighting, color editing, etc.
Visual recognition is affected by changes in visual appearance (pose, scale, rotation) that do not affect a scene’s semantic category. We introduce a method for using sets of traNational Science Foundationorming examples, to learn representations that are robust to these changes.
Spatio-temporal convolutional networks are a good model of how visual cortex represents the actions of others, and thinking about robustness to complex transformations, is key to uncovering how human visual cortex is organized.
We study the problem of 3D object generation. We propose a novel framework, 3D Generative Adversarial Network (3D-GAN), leveraging recent advances in volumetric convolutional networks and generative adversarial nets.
We develop statistical models that are prescriptive rather than predictive/descriptive. From an observational dataset, our methods learn to automatically identify beneficial actions that will improve outcomes, rather than requiring human-made decisions.
We are interested in applying insights from distributed computing theory to understand how ants and other social insects work together to perform complex tasks such as foraging for food, allocating tasks to workers, and choosing high quality nest sites.
Developed at MIT’s Computer Science and Artificial Intelligence Laboratory, a team of robots can self-assemble to form different structures with applications in inspection, disaster response, and manufacturing
On October 16, 2019, Prof. David Patterson, UC Berkeley professor emeritus, Google distinguished engineer, and RISC-V Foundation vice-chair, gave a Dertouzos Distinguished Lecture at CSAIL / MIT, entitled 'Domain Specific Architectures for Deep Neural Networks: Three Generations of Tensor Processing Units (TPUs).'
Septmeber 18, 2019 - Prof. Yoshua Bengio, Prof. University of Montreal, and Scientific Director, Mila, gave a Dertouzos Distinguished Lecture at CSAIL entitled 'Learning High-Level Representations for Agents'
The challenge that motivates the ANA group is to foster a healthy future for the Internet. The interplay of private sector investment, public sector regulation and public interest advocacy, as well as the global diversity in drivers and aspirations, makes for an uncertain future.
We focus on finding novel approaches to improve the performance of modern computer systems without unduly increasing the complexity faced by application developers, compiler writers, or computer architects.
Our interests span quantum complexity theory, barriers to solving P versus NP, theoretical computer science with a focus on probabilistically checkable proofs (PCP), pseudo-randomness, coding theory, and algorithms.
We propose a novel aspect-augmented adversarial network for cross-aspect and cross-domain adaptation tasks. The effectiveness of our approach suggests the potential application of adversarial networks to a broader range of NLP tasks for improved representation learning, such as machine translation and language generation.
Using AI methods, we are developing an attack tree generator that automatically enumerates cyberattack vectors for industrial control systems in critical infrastructure (electric grids, water networks and transportation systems). The generator can quickly assess cyber risk for a system at scale.
Automatic speech recognition (ASR) has been a grand challenge machine learning problem for decades. Our ongoing research in this area examines the use of deep learning models for distant and noisy recording conditions, multilingual, and low-resource scenarios.
We aim to base a variety of cryptographic primitives on complexity theoretic assumptions. We focus on the assumption that there exist highly structured problems --- admitting so called "zero-knowledge" protocols --- that are nevertheless hard to compute
We study the fundamentals of Bayesian optimization and develop efficient Bayesian optimization methods for global optimization of expensive black-box functions originated from a range of different applications.
BlueDBM is an architecture of computer clusters consisting of fast distributed flash storage and in-storage accelerators, which often outperforms larger and more expensive clusters in applications such as graph analytics.
We are working on methods to analyze and process 3D shapes from representations of their boundaries; we focus on extrinsic geometry, that is, how the surface curves and bends through surrounding space.
We aim to study the impact of computer-supported roleplaying in changing social perspectives of digital media users. Such media could take the form of videogames, VR systems, training software, and other types of interactive narrative technology.