DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING
UNIVERSITY OF CALIFORNIA, SAN DIEGO

CSE 291: Hyperdimensional Computing

Winter 2022

Instructors:

Tajana Rosing
Email: tajana.teach at end.ucsd.edu
Office hours (zoom): F 2:00-3:00pm and by appointment

Anthony Thomas
Email: ahthomas at eng.ucsd.edu
Office hours: by appointment

Note: when emailing the instructors with questions about the class, please put "CSE 291" in the subject line.

Meeting time: W 3:00-4:30 on Zoom; link provided on Canvas

Course description:
This course is intended to introduce students to the theory and practice of Hyperdimensional Computing (HDC), an area of research at the intersection of theoretical neuroscience, machine learning, and low-power hardware design. The course will cover the theoretical underpinnings of HDC as well as implementation in hardware, and will consist of a mix of lectures by faculty and invited speakers, followed by the presentations by students. Students who wish a letter grade in the course will need to complete a project related to HDC and turn in a 6-page conference-style paper describing the project. The course will be held via zoom. All attendees agree to being recorded by the virtue of being present in the class.

Grading:

Presentations: All students will be expected to complete an in-class presentation of one or more peer-reviewed research papers related to HD computing. For a list of possible papers see: here. Presentations will be either individual or in groups depending on enrollment. Presentations may consist of one longer paper, or a few related shorter papers. Student presentations will take approximately the last 30 minutes of class time, and may be divided between multiple students depending on the enrollment. Presentations should clearly address the questions listed in How to Read an Engineering Research Paper provided by Professor Bill Griswold. Additionally, all enrolled students should provide feedback on the presentations by filling out a google form. Feedback will be due within 24hrs of the day of presentation, and will be made available to the presenter anonymously.

A prioritized list of papers you would like to present should be submitted to the instructors via Google form here by January 10th at 8am.

Course Project: Students enrolled for a letter grade will be expected to submit a six-page paper in 2-column IEEE conference format by the end of the day Friday, March 11th, via Canvas. The paper should satisfy one of the following:

  1. A critical review of several related papers in HD computing. The paper should synthesize the key contributions and methods of each paper. What do they add to the field? What do they miss and what more needs to be done? The paper should function as a shorter versin of UCSD CSE's research exam. See here for more information.
  2. An original research idea of your own choosing.
Project proposals should be submitted to the instructors via the Google form here by January 10th at 8am.

Schedule (subject to change)

Date Topic Recommended Reading Student Presenters Presenter Feedback Form
Jan 5 Introduction and Course Organization (Tajana Rosing)
Theory I: Encoding and decoding of discrete data, probability tools, noise robustness: Anthony Thomas, UCSD
Jan 12 Introduction to Statistical Learning: Guest lecture by Sanjoy Dasgputa, UCSD
Sparse Random Representations in Neuroscience: Guest lecture by Yonatan Aljadeff, UCSD 		Emergence of Simple-Cell Receptive Field Properties by Learning a Sparse Code for Natural Images [pdf]
Jan 19 Introduction to Computing in Distributed Representation with High-Dimensional Vectors: Guest Lecture by Pentti Kanerva, RCTN, UC Berkeley Hyperdimensional Computing: an Introduction to Computing in Distributed Representation with High-Dimensional Vectors [pdf]
Holographic Reduced Representations [pdf]
Jan 26 Theory II: Encoding for continuous data and learning in HDC: Anthony Thomas, UCSD
Functional VSAs: Guest Lecture by Chris Kymm, RCTN, UC Berkeley 		A Theoretical Perspective on Hyperdimensional Computing [pdf]
Random Features for Large Scale Kernel Machines [pdf, blog post]
Computing on Functions Using Randomized Vector Representations [pdf]
Feb 2 Hyperdimensional Computing and Neural Networks: Guest Lecture by Denis Kleyko, RCTN, UC Berkeley
Resonator Networks: Guest Lecture by Paxon Frady, RCTN, UC Berkeley		 Resonator Networks for Factoring Distributed Representations of Data Structures [pdf]
Density Encoding Enables Resource-Efficient Randomly Connected Neural Networks [pdf]
Feb 9 Classification, clustering, and other HDC Applications: Guest Lecture by Yeseong Kim, DGIST
Hashing and HDC: Anthony Thomas, UCSD		 Network Applications of Bloom Filters: A Survey [pdf]
Efficient Biosignal Processing Using Hyperdimensional Computing: Network Templates for Combined Learning and Classification of ExG Signals [pdf ]
Feb 16 Vision, Robotics, Symbolic Reasoning and HD Computing: Guest lecture by Yiannis Aloimonos, UMD Learning Sensorimotor Control with Neuromorphic Sensors: Toward Hyperdimensional Active Perception [pdf]
Symbolic Representation and Learning with Hyperdimensional Computing [pdf]
Feb 23 Hardware Implementations of HD Computing: Guest lecture from Mohsen Imani, UCI Vector Symbolic Architectures as a Computing Framework for Nanoscale Hardware [pdf]
FACH: FPGA-based Acceleration of Hyperdimensional Computing by Reducing Computational Complexity [pdf]
Mar 2 Security, Privacy, and Adversarial learning: Guest lecture by Xun Jiao, Villanova PoisonHD: Poison Attack on Brain-Inspired Hyperdimensional Computing [pdf]
HDTest: Differential Fuzz Testing of Brain-Inspired Hyperdimensional Computing [pdf
Energy-Efficient Brain-Inspired Hyperdimensional Computing Using Voltage Scaling [pdf]
Mar 9 Student Presentations and Conclusion