Asynchronous and adaptive stochastic approximation methods for accelerating deep learning
Stochastic approximation methods (SAMs) have been extensively applied in deep learning.
Learning and Embedding Ethical Guidelines in Group Decision-Making AI
In many situations, a group of people needs to make a joint decision.
Neural Memories for Text and Knowledge Graphs
Neural memories based on distributed representations within neural networks offer the ability of robust retrieval in the presence of noisy, partial and approximate inputs.
Improving Generalization and Abstraction in Deep Reinforcement Learning
This project seeks to develop novel methods for improving the generalization of learning within the context of computational reinforcement learning, by leveraging theoretical results from cognitive science research.
Fast Learning of Neural Network Models with Provable Generalizability
The parameters of a deep neural network are learned from the training data by minimizing a nonconvex empirical risk function.
AI Models for Curation of Threat Intelligence
Security threat intelligence data is scattered, has high volume, and comes in a variety of formats.
Active Learning of Adversarial Attack Boudaries
This research program introduces a novel framework for generating adversarial examples for machine learning (ML) algorithms.
Extracting Types from Python Machine Learning Libraries
Our work aims to develop novel AI Automation techniques and tools that (1) improve productivity of AI/ML researchers and practitioners and (2) improve correctness, robustness, and maintainability of AI/ML software.
A Code Knowledge Graph for Planning Data Science Experiments
Semantic shift as measure of bias with applications to detection, explanation and mitigation of misinformation
We develop methods to quantify semantic shift between two comparable text collections as a way to identify the concepts with similar and different semantic neighborhoods.
Manifold-Structured Latent Space for Deep Generative Modeling
Exploring the manifold structure of the latent space by using multiple charts and their transition functions, a novel enhancement of current generative models whose latent space is modeled with a single-chart structure.
MIKA: Music Knowledge Intelligence Agent
Machine learning systems coupling symbolic- and signal-based inputs have been almost exclusively used for speech recognition MIKA extends this work to more complex signals, testing on a historical corpus of jazz.
Data Recovery and Subspace Clustering from Quantized and Corrupted Measurements
New subclustering methods to recover data from degraded or corrupted data to improve image and video processing, network analysis, and data privacy applications.
Neural Memories: Distributed Representations and Associative Retrieval
Deep neural networks may offer a new, more robust alternative to traditional computer memories in the presence of noisy data.
Capacity Limited Reinforcement Learning in Minds and Machines
Exploit the cognitive science of “generalization” in human learning in order to develop improved AI algorithms.
Exploration of Artificial Intelligence Approaches to Earth Observing Remote Sensing
Using AI techniques to improve remote sensing, with a focus on assessing water quality at a global spatial scale.
Tentacular AI (TAI)
TAI creates a community of knowledge-based agents, tied to the Internet of Things, which can reach out to each other when a problem exceeds their individual capabilities.