Recent Research Awards: Highlights

 

Project Summary 

This project seeks to a) adapt large-scale foundry-ready synthesis processes for 2D transition metal disulfides and -tellurides to the specific requirements of photonic device structures, b) explore TMD heteroepitaxy on amorphous patterned substrates for preparation of monolithic structures including strained films, and combine these resulting quality films with c) strainoptronics device concepts to enhance photoabsorption in elsewise unattainable frequency bands, d) prototype figure-of-merit engineered optoelectronic devices based on synergistic combinations of heterojunctions and geometric setup for enhanced light-matter-interactions, and e) demonstrate optical gain by monolithic integration of 2D material films into photonic integrated circuits (PIC). Our work will use foundry-based commercial tape-out or foundry-near waferscale substrates and seeks to avoid any non-scalable techniques such as transfer and exfoliation.

Our novel synthesis capabilities paired with figure-of-merit enhanced optoelectronic device prototypes bears relevance for the USAF, general military, but also civilian applications since PICs, detectors, and modulators are integral components for the ever-increasing data communication and processing needs, especially in light of emerging photonic intelligence information processing such as in artificial photonic neural networks for machine learning tasks.  

 

Project Summary

Modern computing hardware is constrained by stringent requirements like extremely small size, low power consumption, and high reliability. Consequently, unconventional computing methods such as Stochastic Computing (SC) that directly address these issues, are of increasing interest, especially for Machine Learning (ML) applications in Artificial Intelligence (AI). SC is a novel computation framework in which input data is continuously provided by streams of bits; therefore, complex computations can then be computed by simple bit-wise operations on the streams. The main attraction of SC is that it enables very low-cost and low-power architectural implementations, especially for arithmetic operations using simple logic elements. This feature is very relevant to Neural Networks (NNs), because NNs require significant hardware resources, therefore consuming substantial power when computing big data for ML; moreover, current NN architectures are difficult to configure to suit different applications, because the hardware is rather complex and not very flexible. So as ML systems are reaching the fundamental limits of computation using NNs, SC has emerged as a plausible and practical solution to meet performance, energy and resilience requirements for massive parallelism and fast deployment of hardware to support AI with direct impact on technology and national economic growth. The goal of this project is to develop NN architectures that rely on different computational features for cross-cutting schemes (spanning hardware units, algorithms, and applications) aimed at designing such efficient SC-based NNs. 

The technical work pursued under this project exploits the main features of SC and proposes a sound research program with several novel concepts. The first novelty of this investigation is that it makes possible the design of SC NNs by focusing on architectural level hardware targeting also important metrics for SC (such as reducing latency and improving accuracy, mostly in inference and training). The second novelty of this work is that it addresses fundamental issues in which simple SC hardware is utilized adaptively to data to sustain a high level of parallel computation in NNs; solutions revolve around a configurable bottom-up scheme in which initially low-level hardware (such as neurons and processing function units) are modularly employed in the NNs to support computation at higher levels. Novel memory organizations to remedy errors when SC is employed are also proposed; this also enhances application-dependent requirements. The third novelty is the provision of having both SC as well as conventional (binary) computation on one combined hardware implementation; this is an added benefit for optimizing computing performance just in case the SC does not meet the accuracy requirements of the application at hand. Therefore, this timely research is directed to the continued technical innovation for emerging computing systems and architectures with relevance to both the computing and ML communities and strong implications on advancements in society and the US computing industry-at-large; moreover, this project is strongly committed to Broadening Participation in Computing (BPC) and its success.

 

Project Summary

The objective of this funded project is to design and assess the analytical performance of the Au Nano sensor from Dr. Zaghloul’s lab that directly detects SARS-CoV-2 virus and distinguishes that binding from other human coronaviruses. The sensor developed is a nanohole array (NHA)-based plasmonic sensing system. It shows high levels of sensitivity and selectivity for detecting and analyzing analytes, especially mixtures of chemicals/biochemicals.  We have filed a joint patent for the technology. 

This project will leverage this patented technology for an infectious disease application. We propose to build microfluidic channel on top of the NHA for sample input and fluid output to detect SARS-CoV-2 virus and distinguish that the binding analyte from other human coronaviruses. We will use open-source machine learning libraries to train the ANN with the images acquired by a smartphone. The purpose of this step is to classify SARS-CoV-2 with more than 90% accuracy.  The device will be designed for in home personnel care.

 

Project Summary

This project sets the stage for leveraging the concurrent and synergistic advances in digital domains such as machine learning, high-performance computing, big data, clouds, and smart and connected devices to create new breads of applications that were hardly possible to conceive and deploy before. This can be productively accomplished by presenting this plethora of different domains to application developers as one integrated system, which is referred to here as the digital continuum. This project creates a testbed infrastructure for such digital continuum with capabilities for application deployments, task distributions, performance monitoring and optimizations, and data sets for experimentation with emphasis on distributed intelligence. The concepts and access to such testbed are shared with the community in order to open new doors for innovative system and application research in the digital continuum as a system.

Convergence research is identified as one of the big ideas for future progress and investments. The digital continuum is an example of such converged systems. This project creates an experimental infrastructure, IDIEA-DC that can mimic the digital continuum from the edge devices to the data center. IDEA-DC also provides productive interfaces for exploring and testing research ideas in support of launching distributed machine learning applications on the digital continuum as a system, such as federated learning. It supports monitoring and control for intelligent and dynamic cross-layer optimizations with respect to many different parameters including performance and bandwidth availability, energy and quality of service. Two instances of such infrastructure will be created. One will be a heavily instrumented cluster devoted to emulate such digital continuum while providing the opportunity for more accurate measurements and closer control. The second will be a cloud-based deployment. The infrastructure will be augmented with relevant data sets, as well as libraries for measurements and control.

 

Project Summary

The project aims to advance the state of the art in learning and data gathering processes, and contribute to the science base of machine learning, control/learning theory and Bayesian statistics. The overarching goal of the project is to develop tools for enabling optimal incorporation of the user information during the learning process and learning from multiple data sources acquired by non-similar users/experts.