ASCENT Benchmarking Liaison Meeting/Using and Evaluating ASCENT Technologies in Support of Few-Shot Machine Learning Models, Homomorphic Encryption Algorithms, and Analog Synapse Circuits

Using and Evaluating ASCENT Technologies in Support of Few-Shot Machine Learning Models, Homomorphic Encryption Algorithms, and Analog Synapse Circuits


Presenters: Arman Kazemi, Ann Franchesca Laguna, Dayane Reis (students) and Ramin Rajaei (post-doc), and Prof. Michael Niemier (Notre Dame)

In this meeting, we will review recent work related to compute-in-memory support for homomorphic encryption schemes and prototypical networks for few-shot learning applications. Initial case studies are purposely SRAM-based in order to establish a baseline as to what performance gains are derived from an architecture, and what performance gains are derived from technology. (While our initial studies are SRAM-based, we will highlight what role ASCENT technology should play in this space, and outline next steps in design and benchmarking in this regard.)  We will also discuss recent design efforts with a hybrid synapse structure comprised of ferroelectric metal field effect transistors (FeMFETs) and will compare this design to FeFET and PCM-based designs.

Tasks: 2776.042, .043, and .044.

Please note, this meeting is only available to the JUMP research community, such as Principal Investigators, Postdoc researchers, Students, and Industry/Government liaisons