This $299,889 federal Project Grant award from the National Science Foundation's (NSF) Mathematical and Physical Sciences (CFDA 47.049) program supports research at the University of California, San Diego (UCSD) to develop algorithms for compressing and improving the efficiency of large neural networks used in modern artificial intelligence applications. The key products and services to be delivered include: The research project focuses on developing quantization, pruning, and low-rank...
This Project Grant award of $299,999 from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) Federal Grant Program supports the development of a cyberinfrastructure that will seamlessly integrate lossy data compression into deep learning pipelines for scientific applications. The key products and services to be delivered include: A user-friendly interface that allows users to define accuracy requirements and instantiate different data compression...
The National Science Foundation (NSF) awarded a $300,000 Project Grant to North Carolina State University (NC State) under the NSF's Computer and Information Science and Engineering (CISE) Federal Grant Program (CFDA 47.070). The purpose of this 3-year grant, effective July 1, 2025 through June 30, 2028, is to develop a cyberinfrastructure that seamlessly and adaptively integrates lossy data compression techniques into deep learning pipelines within scientific applications. This integration aims...
This National Science Foundation (NSF) Project Grant award under the Computer and Information Science and Engineering (CFDA 47.070) program aims to develop software frameworks that can efficiently serve and deploy machine learning models for a variety of AI-powered applications. The $600,000 award, spanning from October 2024 to September 2027, tasks the prime awardee, Georgia Tech Research Corporation, with creating agile mechanisms and policies to serve a family of AI models across...
This Project Grant award from the National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) program aims to advance the efficiency of machine learning model inference through a compression-aware computing framework. The $171,387 project, awarded to the Stevens Institute of Technology, will develop machine learning models capable of self-awareness in response to lossy compression techniques like sparsification and quantization. This will enable the recovery and...
This three-year, $276,611 National Science Foundation Project Grant supports research at the University at Albany to develop algorithms and theory for compressing deep neural networks. Funded through NSF's Mathematical and Physical Sciences program (CFDA 47.049), this award will advance knowledge in discrete optimization and machine learning. Key products include new coarse gradient and thresholding algorithms to enable efficient deployment of AI systems on mobile and low-power platforms. By...
This $134,992 Project Grant award from the National Science Foundation (NSF) Computer and Information Science and Engineering (CISE) program aims to develop methods and architectural support for energy-efficient and scalable artificial intelligence (AI) systems. The key objectives are to: Leverage dynamic connectivity to reduce redundancy in AI models by adapting them to specific tasks and data. Provide architectural support for elastic processing through heterogeneous architectures...
This Project Grant award of $600,000 from the National Science Foundation's (NSF) Computer and Information Science and Engineering (CISE) program (CFDA 47.070) supports research at the University of Central Florida (UCF) to develop software-hardware solutions that enable efficient inference of large AI foundation models on resource-limited computer systems. The key research thrusts include: 1) designing a sparsity-aware scheduling system to address data transfer bottlenecks between GPUs and...
This National Science Foundation (NSF) Technology, Innovation, and Partnerships (CFDA 47.084) Project Grant award of $274,623 to Ai Acuity, a for-profit organization, aims to develop improved image compression technology that targets machine learning-based detection algorithms. The key objective is to reduce the size of Earth observation imagery, thereby lowering the costs associated with transmission, storage, and processing. This technology will also have applications in high-efficiency...
This National Science Foundation (NSF) Project Grant award for $599,999 supports research at Cornell University from June 15, 2023 to May 31, 2026. The objective is to develop a "modern theory of data compression" to explain the performance of artificial neural network-based compression algorithms and identify avenues for future improvements. The research aims to advance the field of data compression, which has important implications for enabling more realistic and immersive...
This $199,324 Project Grant awarded by the National Science Foundation's Engineering program (CFDA 47.041) to Kennesaw State University Research And Service Foundation, Inc. is funding a research project to develop a novel compression technique called Predictor to Prefetcher (P2P) that leverages activation sparsity within large AI models. The goal is to enable more effective and efficient deployment of large AI/ML models on resource-constrained edge devices. The project will design, implement, and verify the P2P approach, starting with an analysis of predictability in feed-forward network activation patterns. It will also investigate building a device-end predictor using a tiny ML model. Additionally, the project will extend the P2P approach to large vision and multimodal models, and create an open-source simulator to assess its effectiveness in reducing cache/memory pollution and execution times. This research aims to expand knowledge in AI science and education, supporting the ongoing growth of large AI models and maintaining US leadership in AI technology.