Yiwei Lu

I am a final year Ph.D. student in the David R. Cheriton School of Computer Science at the University of Waterloo, where I am supervised by Prof. Yaoliang Yu and Dr. Sun Sun. I am also a student affiliate of the Vector Instituite and a research affliate of The Salon with Prof. Gautam Kamath.

Previously, I have completed my M.Sc. in Computer Science at the University of Manitoba, where I was advised by Prof. Yang Wang. I did my bachelors at the University of Electronic Science and Technology of China. I was also an exchange student at UC Santa Barbara.

I am currently on the academic job market for faculty positions.

Email  /  CV  /  Google Scholar  /  Github  /  LinkedIn

  New! July 2024: New paper on arXiv: Machine Unlearning Fails to Remove Data Poisoning Attacks.
  New! June 2024: New paper on arXiv: Alignment Calibration: Machine Unlearning for Contrastive Learning under Auditing.
  May 2024: One paper accepted to ICML 2024: Disguised Copyright Infringement of Latent Diffusion Models.
  Feb 2024: One paper accepted to SaTML 2024 (presented on April 11th).
  Jan 2024: I am a winner of the Cheriton Scholarship.
  Dec 2023: I am selected as a top reviewer in NeurIPS 2023.
  Oct 2023:One paper accepted to NeurIPS 2023.
  Oct 2023:One paper accepted to TMLR.
  May 2023: One paper accepted to ICML 2023.
  Dec 2022: One paper accepted to TMLR.
  Feb 2022: I am joining Huawei Noah's Ark Lab in Montreal as an Intern.
  January 2021: I am joining National Research Council of Canada as an Intern
  August 2020: I am joining University of Waterloo and Vector Instituite as a Ph.D. student

My research focuses on trustworthy machine learning, specifically examining how external training data affects model performance and robustness. This includes studying data poisoning attacks, the impact of problematic training data (e.g., disguised copyrighted material), and developing machine unlearning techniques to mitigate their effects.

More generally, I am interested in building a trustworthy machine learning pipeline, spanning from data to training procedures and models. This includes topics in (but is not limited to) memorization, data attribution, neural network quantization, self-supervised learning, diffusion training, and tools in non-convex optimization.

We address machine unlearning for contrastive learning pretraining schemes via a novel method called Alignment Calibration. We also propose new auditing tools for data owners to easily validate the effect of unlearning.

We find that current approaches for machine unlearning (MUL) are not effective at removing the effect of data poisoning attacks.

We find that neural network quantization offers improved robustness against different data poisoning attacks.

We reveal the threat of disguised copyright infringement of latent diffusion models, where one constructs a disguise that looks drastically different from the copyrighted sample yet still induces the effect of training Latent Diffusion Models on it.

We study indiscriminate data poisoning attacks against pre-trained feature extractors for fine-tuning and transfer learning tasks and propose feature targeted attacks to address optimization difficulty under constraints.

We propose forward backward proximal quantizers for understanding approximate gradients in neural network quantization and provide a a new tool for designing new algorithms.

We propose a general and novel loss function on contrastive learning based on f-mutual information. Additionally, we propose a f-Gaussain similarity funcntion with better interpretability and empirical performance.

We propose CM-GAN by combining the main strengths of diffusions and GANs while mitigating their major drawbacks.

We find (1) existing indiscriminate attacks are not well-designed (or optimized), and we reduce the performance gap with a new attack; (2) there exists some intrinsic barriers of data poisoning attacks, namely that when the poisoning fraction is smaller than a (easy to calculate) threshold, no attack succeeds.

We find that neural networks are surprisingly hard to (indiscriminate) poison and give better attacks.

We propose a general and novel loss function on contrastive learning based on f-mutual information.

We propose a new problem called unlabeled scene adaptive crowd counting.

We propose a more realistic problem setting for anomaly detection in surveillance videos and solve it using a meta-learning based algorithm.

We propose a structure learning framework that retains the pairwise similarities between the data points.

We propose a novel sequential generative model based on variational autoencoder (VAE) for future frame prediction with convolutional LSTM (ConvLSTM).

We propose a novel similarity learning framework by minimizing the reconstruction error of kernel matrices, rather than the reconstruction error of original data adopted by existing work.

We propose a ConvLSTM-based model to perform semantic segmentation on compressed videos directly. This significantly speed up the training and test speed.

Anomaly Detection in Surveillance Videos using Deep Learning - Yiwei Lu, M.Sc.thesis, Department of Computer Science, University of Manitoba, June 2020.

Credits to Jon Barron for the website design.