Yefan Zhou

Hi, I'm Yefan. I am a first-year CS PhD student at Dartmouth College advised by Prof. Yaoqing Yang. I am a researcher at UC Berkeley/ICSI working with Prof. Michael Mahoney.
I earned my Master's degree in EECS at UC Berkeley.

Scholar  /   /   /  Linkedin  /  Email  /  Resume

I'm interested in improving transparency and efficiency of machine learning models. My current research is focused on model diagnosis, utilizing high-dimension features such as loss landscapes, weight matrix analysis. This research contributes to practical ML development, including neural network training, model compression and hyperparameter tuning.

Most deep neural networks have complex multilayer structures, often seen as a barrier to transparency. In our research, we reveal a significant insight: these layers are not uniformly well-trained. We introduce a new training method that utilizes "model diagnostic" tool to identify and address underperforming layers, and enhance the overall network quality.

The study identifies a transition phenomenon in neural network pruning, where the effect of increasing the temperature-like parameter (e.g. training epochs) depends on the value of the load-like parameter (e.g. pruning ratio), leading to different pruning outcomes. The findings are then applied to three practical scenarios, including optimizing hyperparameters for improved pruning and selecting the most suitable model for pruning.

A SVR model can be disposed towards recognition (classification-based) or reconstruction depending on how dispersed the training data becomes.
We propose "dispersion score", which is a data-driven metric used to measure the tendency of SVR models to perform recognition or reconstruction. It can also be used to diagnose problems from the training data and guide the design of data augmentation schemes.

Empirical grasping datasets are typically sparsely labeled (i.e., a small number of successful grasp labels in each image).
We propose a maximum likelihood grasp sampling loss (MLGSL) for learning robotic grasping from sparsely labeled datasets.
MLGSL is 8× more data-efficient than SOTA with a 91.8% grasp success rate in real-world experiments.

ICML 2024, CAPL 2024, ICLR 2024, CVPR 2024, NeurIPS 2023, IROS 2022

We proposed Hybrid Policy Gradient (H-PG), a novel deep reinforcement learning framework for robotic grasping task defined in continuous-discrete hybrid action space;
H-PG improves baseline by 7.4% of grasp success rate on YCB dataset in PyBullet simulator.