Sunny Y. Duan

Sunny Y. Duan

PhD Candidate in Brain and Cognitive Sciences

sunnyd@mit.edu

Massachusetts Institute of Technology

I am a PhD candidate in Brain and Cognitive Sciences at MIT, working with Prof. Ila Fiete in the Fiete Lab. My research combines computational neuroscience with machine learning to understand how the brain works. I am interested in how the circuitry of the brain gives rise to computation and behavior in two broad ares:

Neural Circuit Modeling

I develop dynamical systems models of neural circuits in Drosophila and C. elegans to understand how neural dynamics give rise to behavior. My work focuses on making minimal, biologically plausible modifications to connectome data to produce functional circuits.

Multi-area Motor Control

I study how different brain regions (cerebellum, motor cortex, and sensorimotor cortex) cooperate to enable versatile motor learning. Using neural network models, I investigate how these systems handle real-world challenges like delayed feedback and noisy sensory input.


Prior to MIT, I worked as a Research Engineer at DeepMind, where I developed reinforcement learning algorithms for data center power optimization. I also contributed to the Aquila project at Google.

Selected Publications

Uncovering Latent Memories: Assessing Data Leakage and Memorization Patterns in Large Language Models

S Duan, M Khona, A Iyer, R Schaeffer, IR Fiete

arXiv preprint arXiv:2406.14549, 2024

Representation learning for improved interpretability and classification accuracy of clinical factors from EEG

Garrett Honke, Irina Higgins, Nina Thigpen, Vladimir Miskovic, Katie Link, Sunny Duan, Pramod Gupta, Julia Klawohn, Greg Hajcak

arXiv preprint arXiv:2010.15274, 2022

Unsupervised model selection for variational disentangled representation learning

Sunny Duan, Loic Matthey, Andre Saraiva, Nicholas Watters, Christopher P Burgess, Alexander Lerchner, Irina Higgins

arXiv preprint arXiv:1905.12614, 2019

Users really do plug in USB drives they find

M Tischer, Z Durumeric, S Foster, S Duan, A Mori, E Bursztein, M Bailey

2016 IEEE Symposium on Security and Privacy (SP), 306-319

See and Copy: Generation of complex compositional movements from modular and geometric RNN representations

S Duan, M Khona, A Bertagnoli, S Chandra, IR Fiete

NeurIPS Workshop on Symmetry and Geometry in Neural Representations, 388-400

Education

Massachusetts Institute of Technology

PhD Candidate in Brain and Cognitive Sciences

GPA: 4.00/4.00

Advised by Prof. Ila Fiete

Research: Nonlinear dynamical systems inference for c-elegans brain modeling

University of Illinois at Urbana-Champaign

Bachelor of Science in Computer Engineering

GPA: 3.83/4.00

Honors: Magna Cum Laude; Dean's List (5 semesters)

Professional Experience

Ghost Locomotion

Software Engineer

Developed perception models for self-driving cars; work utilized to successfully deliver first autonomous road test

DeepMind

Research Engineer

April 2018 – April 2020

  • Implemented reinforcement learning algorithms for power optimization
  • Led research initiative using learned simulation for analysis and sim-to-real applications
  • Tech lead for YouTube promotions optimization using TD-learning
  • Developed novel method for estimating disentanglement in VAE models

Google

Software Engineer III

August 2016 – April 2018

  • Led engineering initiative for ASIC switches networking protocol
  • Implemented cloud management software for controlling networking switches