Deep Learning Master Class by Dr. Agostinelli

Forest Agostinelli gave a Master Class at the 17th International Symposium on Combinatorial Search (SoCS) on Deep Learning, Reinforcement Learning, and Heuristic Search. The one-hour talk covered his research group’s advancements in the combination of machine learning and heuristic search and its application to problems such as the Rubik’s cube, quantum algorithm compilation, and reaction mechanism prediction.

A Better Assessment provides a quick history of Dr. Agostinelli's work.

First Place Award at DAC University Demonstration

The iCAS lab, directed by Dr. Ramtin Zand, won the first-place award at the 2024 University Demonstration at DAC, The Chips to Systems Conference for the project titled "HiRISE: High-Resolution Image Scaling for Edge ML via In-Sensor Compression and Selective ROI."

Demo by: Brendan Reidy and Peyton Chandarana, Ph.D. Research Assistants

Project Description: With the rise of tiny IoT devices powered by machine learning (ML), many researchers have directed their focus toward compressing models to fit on tiny edge devices. Recent works have achieved remarkable success in compressing ML models for object detection and image classification on microcontrollers with small memory, e.g., 512kB SRAM. However, there remain many challenges prohibiting the deployment of ML systems that require high-resolution images. Due to fundamental limits in memory capacity for tiny IoT devices, it may be physically impossible to store large images without external hardware. To this end, we propose a high-resolution image scaling system for edge ML, called HiRISE, which is equipped with selective region-of-interest (ROI) capability leveraging analog in-sensor image scaling. Our methodology not only significantly reduces the peak memory requirements, but also achieves up to 17.7x reduction in data transfer and energy consumption.

Paper Authors: Brendan Reidy, Sepehr Tabrizchi, MohammadReza Mohammadi, Shaahin Angizi, Arman Roohi, and Ramtin Zand

Jamshidi Earns Recognition for Most Influential Paper

Jamshidi received the Most Influential Paper Award in April at the 19th International Conference on Software Engineering for Adaptive and Self-Managing Systems (SEAMS) in Lisbon, Portugal. Jamshidi’s paper, “Autonomic Resource Provision for Cloud-based Software,” was submitted, accepted and published just prior to earning his Ph.D. from Dublin City University in Ireland in 2014. It was presented at the 2014 SEAMS Conference in India. See original post for details.

Narayanan Seeks to Enhance Safety, Efficiency of Dynamic Systems through AI

Since 2021, Assistant Professor Vignesh Narayanan has taught in the Department of Computer Science and Engineering and is affiliated with the Artificial Intelligence Institute of the University of South Carolina (AIISC) and Carolina Autism and Neurodevelopment (CAN) Research Center. He is passionate about the integration between AI and dynamic systems, and its impact on safety and efficiency for consumers. Narayanan’s research surrounds the interaction between humans and dynamic systems to prevent such systems from unsafe behavior as they change over time. 

Read the full article here.

Dr. Sur Receives NSF Research Award

Dr Sanjib Sur has received a National Science Foundation (NSF) research award for his project titled "Modernizing Underground Mines Operations with Millimeter-Wave Imaging and Networking". You can learn more about his research here and in the article Improving underground mining safety with millimeter wave technology.

"The project aims to address the unique challenges of sensing and networking in underground mining environments by employing millimeter-wave (mmWave) wireless, a core technology for 5G and beyond standards. This technology is particularly suited for the harsh conditions of underground mines, such as dust and low light or dark conditions. However, the adoption of mmWave technology in mining is challenging due to reconstructing high-quality 3D maps in complex structures, fusing static and mobile underground real-time maps, and deploying mmWave communication infrastructures. By overcoming these challenges, this project seeks to enhance safety in mining operations, improve operational efficiency through better resource management, navigation, and machinery positioning, and contribute to the national interest by advancing the future of autonomous mining systems."

Dr. Jamshidi Receives Most Influential Paper Award at SEAMS

We are happy to announce that Dr. Jamshidi, A. Ahmad, and C. Pahl were recipients of a Most Influental Paper Award at the 9th International Symposium of Software Engineering for Adaptive and Self-Managing Systems for their paper "Autonomic Resource Provisioning for Cloud-Based Software."

Students Present Papers at AAAI-MAKE Spring Symposium

From left to right: Amit Sheth, Nicholas Harvel, Dr. Edward Feigenbaum,  Dr. Manas Gaur (USC alumnus), Kaushik Roy, and Yuxin Xi.  Kaushik and Xi are USC CS PhD students.

Students presented the papers below at the  Empowering Machine Learning and Large Language Models with Domain and Commonsense Knowledge (AAAI-MAKE 2024) AAAI Spring Symposium at Stanford University. Download papers here.

  • Yuxin Zi, Kaushik Roy, Vignesh Narayanan, and Amit Sheth presented their paper titled "Exploring Alternative Approaches to Language Modeling for Learning from Data and Knowledge"
  • Kanak Raj, Kaushik Roy, Vamshi Bonagiri, Priyanshul Govil and Krishnaprasad Thirunarayanan:
    "K-PERM: Personalized Response Generation Using Dynamic Knowledge Retrieval and Persona-Adaptive Queries".
  • Kaushik Roy, Alessandro Oltramari, Yuxin Zi, Chathurangi Shyalika, Vignesh Narayanan and Amit Sheth:
    "Causal Event Graph-Guided Language-based Spatiotemporal Question Answering"

Dr. Zand Receives NSF CAREER Award

Ramtin Zand

We are proud to announce that Dr. Ramtin Zand has received an NSF CAREER award for his research on "Heterogeneous Neuromorphic and Edge Computing Systems for Realtime Machine Learning Technologies". 

This project aims to harness the combined capabilities of neuromorphic and edge computing to forge a heterogeneous machine learning system. Its primary goal is to enable computer vision and language models on resource- and energy-constrained devices at an unprecedented scale. It focuses on several key aspects: (1) developing hybrid models that merge the energy efficiency, temporal sparsity, and spatiotemporal processing of spiking neural networks with the global processing of transformer models for complex large-scale computer vision tasks, (2) creating a methodology to deploy large language models on edge devices by employing system-level innovations such as computational graph modifications, custom kernels, and mathematical refactoring, (3) designing a flexible edge artificial intelligence (AI) accelerator to overcome hardware limitations hindering real-time implementation of large transformer models at the edge, (4) seamlessly integrating a heterogeneous system of mobile processors, edge AI accelerators, and neuromorphic hardware for a comprehensive end-to-end solution. Throughout the project, rigorous investigation delves into critical trade-offs between bandwidth, accuracy, performance, and energy consumption.

Also see this article about Dr. Zand and his research.

AAAI Best Demo Award

PosterThe paper titled  "Expressive and Flexible Simulation of Information Spread Strategies in Social Networks Using Planning," by Bharath Muppasani, Vignesh Narayanan, Biplav Srivastava, Michael N. Huhns, has been selected for the Best Demo Award at AAAI-24. AAAI is a top AI conference and was held over the past week.

The work enables detailed simulations of opinion evolution and strategic interventions using planning. Designed to enhance human-AI collaboration, the framework supports the creation of strategies that facilitate a deeper understanding and informed engagement with the opinion evolution in networks. It was selected from 30 demos, which themselves were selected from a pool of 97 submissions. You can read the poster and watch the video presentation.