Research Projects
Today, machine learning is transforming many modern artificial intelligence (AI) applications such as computer vision, speech recognition, machine translation, robot control, medical diagnosis, and news recommender system. However, modern computers, which are based on von Neumann architecture, are extremely inefficient to carry on those AI tasks. As powerful as human brains, biologically-inspired computers have the potential to perform machine learning algorithms faster and consumes less power than traditional von Neumann architectures. Accordingly, neuromorphic system has become the promising computing paradigm to break through the bottleneck of traditional computing systems. In this RET site, teachers will work on discovery-based research projects and obtain transdisciplinary research experience on biologically-inspired computing systems spanning application (cancer detection), algorithm (Spiking Neural Networks), architecture and circuit (synaptic memory design), and device (memristor). This program will provide the teachers a holistic ground for developing creative curricula modules and materials in mathematics, physics, biology, chemistry, engineering, and technology that align well with Next Generation Science Standard (NGSS) and Alabama Course of Study Standards.
Today, machine learning is transforming many modern artificial intelligence (AI) applications such as computer vision, speech recognition, machine translation, robot control, medical diagnosis, and news recommender system. However, modern computers, which are based on von Neumann architecture, are extremely inefficient to carry on those AI tasks. As powerful as human brains, biologically-inspired computers have the potential to perform machine learning algorithms faster and consumes less power than traditional von Neumann architectures. Accordingly, neuromorphic system has become the promising computing paradigm to break through the bottleneck of traditional computing systems. In this RET site, teachers will work on discovery-based research projects and obtain transdisciplinary research experience on biologically-inspired computing systems spanning application (cancer detection), algorithm (Spiking Neural Networks), architecture and circuit (synaptic memory design), and device (memristor). This program will provide the teachers a holistic ground for developing creative curricula modules and materials in mathematics, physics, biology, chemistry, engineering, and technology that align well with Next Generation Science Standard (NGSS) and Alabama Course of Study Standards.
Research Project #1 - Application – Excitation-Scanning Hyperspectral Imaging of Tissues for Cancer Detection (Biology and Chemistry Focused)
Project Mentor: Dr. Silas Leavesley, Full Professor in the Departments of Chemical and Biomolecular Engineering, Pharmacology, and the Center for Lung Biology at USA
In Project #1, the RET participants will work with advanced hyperspectral imaging techniques for early detection of pathological processes in the lung, colon, and other tissues. Hyperspectral imaging collects and processes information from across the electromagnetic spectrum to obtain information that may be used to measure live-cell signaling and whole-tissue physiology. Recently, Dr. Leavesley’s team has developed an excitation-scanning hyperspectral imaging microscope that overcomes the limitations of traditional emission scanning by providing high transmission with short acquisition times. The objective of this project is to test the hypothesis that the hyperspectral imaging data quality across a specified type of tissues can be significantly increased by using a structured search for the optimal microscope parameters.
Faculty research area: Dr. Leavesley is a Full Professor in the Departments of Chemical and Biomolecular Engineering and Pharmacology, as well as the Center for Lung Biology at USA. His current research focuses on developing new imaging technologies and approaches that merge the fields of microscopy, endoscopy, and spectral imaging. The work of studying tissue composition of autofluorescent molecules using hyperspectral imaging directly ties in with and supports ongoing research efforts in Dr. Leavesley’s research lab. His team has already performed a range of initial studies to develop new hyperspectral imaging technologies and to understand what characteristic spectroscopic signatures may be associated with different tissue types. This project focuses on a key area centered around understanding the molecular composition of tissues and whether changes in molecular composition, as estimated through hyperspectral imaging measurements, can be used for discriminating tissue types and cancerous from “normal” tissues.
Learn more about Professor Leavesley on his faculty page.
Project Mentor: Dr. Silas Leavesley, Full Professor in the Departments of Chemical and Biomolecular Engineering, Pharmacology, and the Center for Lung Biology at USA
In Project #1, the RET participants will work with advanced hyperspectral imaging techniques for early detection of pathological processes in the lung, colon, and other tissues. Hyperspectral imaging collects and processes information from across the electromagnetic spectrum to obtain information that may be used to measure live-cell signaling and whole-tissue physiology. Recently, Dr. Leavesley’s team has developed an excitation-scanning hyperspectral imaging microscope that overcomes the limitations of traditional emission scanning by providing high transmission with short acquisition times. The objective of this project is to test the hypothesis that the hyperspectral imaging data quality across a specified type of tissues can be significantly increased by using a structured search for the optimal microscope parameters.
Faculty research area: Dr. Leavesley is a Full Professor in the Departments of Chemical and Biomolecular Engineering and Pharmacology, as well as the Center for Lung Biology at USA. His current research focuses on developing new imaging technologies and approaches that merge the fields of microscopy, endoscopy, and spectral imaging. The work of studying tissue composition of autofluorescent molecules using hyperspectral imaging directly ties in with and supports ongoing research efforts in Dr. Leavesley’s research lab. His team has already performed a range of initial studies to develop new hyperspectral imaging technologies and to understand what characteristic spectroscopic signatures may be associated with different tissue types. This project focuses on a key area centered around understanding the molecular composition of tissues and whether changes in molecular composition, as estimated through hyperspectral imaging measurements, can be used for discriminating tissue types and cancerous from “normal” tissues.
Learn more about Professor Leavesley on his faculty page.
Research Project #2 - Algorithm – Implementation of Spiking Neural Networks for Cancer Detection (Mathematics and Programming Focused)
Project Mentor: Dr. Jingshan Huang, Full Professor in Computer Science at USA
From the viewpoint of algorithm design and software development, one of the key questions associated with biologically-inspired computing systems is which neural network algorithms/models to use. Spiking Neural Networks (SNNs) have become promising candidates for efficient implementation of deep neural networks for biologically-inspired computing systems. Different from conventional Artificial Neural Networks (ANN), SNNs are inspired by information processing in biology, where sparse and asynchronous binary signals are communicated and processed in a massively parallel fashion. SNNs exhibit many desirable properties including low power consumption and event-driven information processing and they are well suited to meet the requirement of tolerating the drawbacks of modern hardware. This project will provide RET participants with practical implementation experience of using SNNs with applications in cancer detection.
Faculty research area: Dr. Huang is a Full Professor of Computer Science in the School of Computing at USA. His research areas are Artificial Intelligence, Big Data, and Biomedical Informatics with a unifying theme of the cancer detection. He has published over 80 refereed papers in mainstream journals as well as leading conferences. This research project, focusing on applying the new generation neural network – Spiking Neural Networks – to cancer detection, is directly aligned with and supports ongoing research efforts in Dr. Huang’s research lab.
Learn more about Professor Huang on his faculty page.
Research Project #3 - Architecture and Circuit – Power-Efficient Crossbar Synaptic Memory for SNN (Physics and Technology Focused)
Faculty mentor: Dr. Na Gong, W. Nicholson Associate Professor in Electrical and Computer Engineering at USA
A distinctive property of deep learning systems is the requirement for a large on-chip static random-access memories (SRAMs) to store the synaptic weights. To enable low-cost power efficient synaptic memory design, PI Gong’s team stuided synaptic data correlation and association relationship and designed a context-aware memory that can store synaptic weights efficiently with data self-recovery capability in the presence of memory failures. The objective of this project is to design a low-power memory architecture for SNNs. RET participants will learn the working mechanisms of SRAM architecture, operate of design tools, and design a memory for verification.
Faculty research area: Dr. Gong is currently W. Nicholson Associate Professor in Computer Engineering at USA. She earned a doctoral degree of Computer Science and Engineering from the University at Buffalo, SUNY. In the past 10 years, she has worked extensively on energy-efficient computer-hardware design, with a special emphasis on memory systems. She has published over 90 peer-reviewed papers in computer science and engineering journals and conference proceedings.
Learn more about Professor Gong on her faculty page.
Faculty mentor: Dr. Na Gong, W. Nicholson Associate Professor in Electrical and Computer Engineering at USA
A distinctive property of deep learning systems is the requirement for a large on-chip static random-access memories (SRAMs) to store the synaptic weights. To enable low-cost power efficient synaptic memory design, PI Gong’s team stuided synaptic data correlation and association relationship and designed a context-aware memory that can store synaptic weights efficiently with data self-recovery capability in the presence of memory failures. The objective of this project is to design a low-power memory architecture for SNNs. RET participants will learn the working mechanisms of SRAM architecture, operate of design tools, and design a memory for verification.
Faculty research area: Dr. Gong is currently W. Nicholson Associate Professor in Computer Engineering at USA. She earned a doctoral degree of Computer Science and Engineering from the University at Buffalo, SUNY. In the past 10 years, she has worked extensively on energy-efficient computer-hardware design, with a special emphasis on memory systems. She has published over 90 peer-reviewed papers in computer science and engineering journals and conference proceedings.
Learn more about Professor Gong on her faculty page.
Research Project #4 - Device – Mitigating Device Variations of Memristors to Enable Biologically-Inspired Computing (Physics and Technology Focused)
Faculty mentor: Dr. Jinhui Wang, Associate Professor in Electrical and Computer Engineering at USA
As am emerging device technology to enable biologically-inspired computing systems, memristors, with a three-layer structure, exhibit multilevel conductance states by external incentive. In a memristor-based biologically-inspired computing system, memristors are connected as synapses for neuron devices that implement the function of the Sum of Product (SOP), therefore achieving a small area foot-print and high-density structure than the traditional CMOS technology. However, a major challenge is that, memristors are not reliable, potentially suffering from non-ideal device properties such as device to device variation, cycle to cycle variation, and on/off ratio variation. In this project, the teachers will work on the techniques can effectively reduce the impact of the device variations in memristors. They will test the hypothesis that the multiple-cell technique is effective in mitigating device variations of memristors.
Faculty research area: Dr. Wang is an Associate Professor in Electrical Engineering and he has worked on VLSI and neuromorphic system for 15 years. Dr. Wang published over 120 papers in prestigious journals and conferences. His team has developed several novel techniques to mitigate the nonlinear effect of memristors in neuromorphic systems.
Learn more about Professor Wang on her faculty page.
This RET program is supported by the National Science Foundation (NSF) award CNS-#1953544.
Faculty mentor: Dr. Jinhui Wang, Associate Professor in Electrical and Computer Engineering at USA
As am emerging device technology to enable biologically-inspired computing systems, memristors, with a three-layer structure, exhibit multilevel conductance states by external incentive. In a memristor-based biologically-inspired computing system, memristors are connected as synapses for neuron devices that implement the function of the Sum of Product (SOP), therefore achieving a small area foot-print and high-density structure than the traditional CMOS technology. However, a major challenge is that, memristors are not reliable, potentially suffering from non-ideal device properties such as device to device variation, cycle to cycle variation, and on/off ratio variation. In this project, the teachers will work on the techniques can effectively reduce the impact of the device variations in memristors. They will test the hypothesis that the multiple-cell technique is effective in mitigating device variations of memristors.
Faculty research area: Dr. Wang is an Associate Professor in Electrical Engineering and he has worked on VLSI and neuromorphic system for 15 years. Dr. Wang published over 120 papers in prestigious journals and conferences. His team has developed several novel techniques to mitigate the nonlinear effect of memristors in neuromorphic systems.
Learn more about Professor Wang on her faculty page.
This RET program is supported by the National Science Foundation (NSF) award CNS-#1953544.