[ICLR 2021 Spotlight Oral] "Undistillable: Making A Nasty Teacher That CANNOT teach students", Haoyu Ma, Tianlong Chen, Ting-Kuei Hu, Chenyu You, Xiaohui Xie, Zhangyang Wang

Related tags

Deep Learning copyright knowledge-distillation teacher-student privacy-protection undistillable
Overview

Undistillable: Making A Nasty Teacher That CANNOT teach students

License: MIT

"Undistillable: Making A Nasty Teacher That CANNOT teach students"

Haoyu Ma, Tianlong Chen, Ting-Kuei Hu, Chenyu You, Xiaohui Xie, Zhangyang Wang
In ICLR 2021 Spotlight Oral

Overview

  • We propose the concept of Nasty Teacher, a defensive approach to prevent knowledge leaking and unauthorized model cloning through KD without sacrificing performance.
  • We propose a simple yet efficient algorithm, called self-undermining knowledge distillation, to directly build a nasty teacher through self-training, requiring no additional dataset nor auxiliary network.

Prerequisite

We use Pytorch 1.4.0, and CUDA 10.1. You can install them with

conda install pytorch=1.4.0 torchvision=0.5.0 cudatoolkit=10.1 -c pytorch

It should also be applicable to other Pytorch and CUDA versions.

Then install other packages by

pip install -r requirements.txt

Usage

Teacher networks

Step 1: Train a normal teacher network
python train_scratch.py --save_path [XXX]

Here, [XXX] specifies the directory of params.json, which contains all hyperparameters to train a network. We already include all hyperparameters in experiments to reproduce the results in our paper.

For example, normally train a ResNet18 on CIFAR-10

python train_scratch.py --save_path experiments/CIFAR10/baseline/resnet18

After finishing training, you will get training.log, best_model.tar in that directory.

The normal teacher network will serve as the adversarial network for the training of the nasty teacher.

Step 2: Train a nasty teacher network
python train_nasty.py --save_path [XXX]

Again, [XXX] specifies the directory of params.json, which contains the information of adversarial networks and hyperparameters for training.
You need to specify the architecture of adversarial network and its checkpoint in this file.

For example, train a nasty ResNet18

python train_nasty.py --save_path experiments/CIFAR10/kd_nasty_resnet18/nasty_resnet18

Knowledge Distillation for Student networks

You can train a student distilling from normal or nasty teachers by

python train_kd.py --save_path [XXX]

Again, [XXX] specifies the directory of params.json, which contains the information of student networks and teacher networks

For example,

  • train a plain CNN distilling from a nasty ResNet18
python train_kd.py --save_path experiments/CIFAR10/kd_nasty_resnet18/cnn
  • Train a plain CNN distilling from a normal ResNet18
python train_kd.py --save_path experiments/CIFAR10/kd_normal_resnet18/cnn

Citation

@inproceedings{
ma2021undistillable,
title={Undistillable: Making A Nasty Teacher That {\{}CANNOT{\}} teach students},
author={Haoyu Ma and Tianlong Chen and Ting-Kuei Hu and Chenyu You and Xiaohui Xie and Zhangyang Wang},
booktitle={International Conference on Learning Representations},
year={2021},
url={https://openreview.net/forum?id=0zvfm-nZqQs}
}

Acknowledgement

Comments
  • The accuracy of the nasty resnext29 on CIFAR-100.

    The accuracy of the nasty resnext29 on CIFAR-100.

    Thanks for your excellent work! When I use the nasty resnext29 as the teacher, the accuracy of the distilled resnet18, shufflenetV2, and resnext 29 is 73.83, 64.82, and 79.87 respectively. I did not change the json file. I wonder if some parameters need additional settings. Looking for your reply!

    opened by liuhao-lh 2
  • Specifying the number of categories in train_nasty.py

    Specifying the number of categories in train_nasty.py

    Hi, thanks for your excellent work~

    In the train_nasty.py, the 'num_classes' for preresnet are not specified, which will raise errors when use cifar100 dataset.

    opened by xuguodong03 1
  • About the setting.

    About the setting.

    Hi there, thanks for your great work!!! I have a question about the setting of this work. Does this work assume that the 'stealer' can access and only access the logits output of the teacher model? And the model resources like model structure and parameters are invisible?

    opened by kxgong 2
  • The accuracy of resnet family

    The accuracy of resnet family

    Hi, thanks for your excellent work~ Training a baseline model(train_scratch.py) on cifar100, resnet18 achieved a higher accuracy than resnet34 and resnrt50. I have nothing changed. Do you know what is wrong? Looking for your reply!

    opened by swift1988 4
  • The asymmetry of KL divergence.

    The asymmetry of KL divergence.

    Hi, I notice that in the 'train_nasty.py', when the KL divergence is computed, normal teacher's output (output_stu) is regarded as input and nasty teacher's output (output_tch) is regarded as target. However, in general KD, the fixed model (teacher) is usually regarded as the target and the model that needs update is regarded as the input.

    I wonder why you adopt an opposite order in KL loss function. Is there any point here? Thanks!

    opened by xuguodong03 1
Owner
VITA
Visual Informatics Group @ University of Texas at Austin
VITA
GitHub
[ICML 2021] "Graph Contrastive Learning Automated" by Yuning You, Tianlong Chen, Yang Shen, Zhangyang Wang

Graph Contrastive Learning Automated PyTorch implementation for Graph Contrastive Learning Automated [talk] [poster] [appendix] Yuning You, Tianlong C

Shen Lab at Texas A&M University 80 Nov 23, 2022
[CVPR 2021] "The Lottery Tickets Hypothesis for Supervised and Self-supervised Pre-training in Computer Vision Models" Tianlong Chen, Jonathan Frankle, Shiyu Chang, Sijia Liu, Yang Zhang, Michael Carbin, Zhangyang Wang

The Lottery Tickets Hypothesis for Supervised and Self-supervised Pre-training in Computer Vision Models Codes for this paper The Lottery Tickets Hypo

VITA 59 Dec 28, 2022
[ICML 2021] “ Self-Damaging Contrastive Learning”, Ziyu Jiang, Tianlong Chen, Bobak Mortazavi, Zhangyang Wang

Self-Damaging Contrastive Learning Introduction The recent breakthrough achieved by contrastive learning accelerates the pace for deploying unsupervis

VITA 51 Dec 29, 2022
[CVPRW 21] "BNN - BN = ? Training Binary Neural Networks without Batch Normalization", Tianlong Chen, Zhenyu Zhang, Xu Ouyang, Zechun Liu, Zhiqiang Shen, Zhangyang Wang

BNN - BN = ? Training Binary Neural Networks without Batch Normalization Codes for this paper BNN - BN = ? Training Binary Neural Networks without Bat

VITA 40 Dec 30, 2022
[Preprint] "Chasing Sparsity in Vision Transformers: An End-to-End Exploration" by Tianlong Chen, Yu Cheng, Zhe Gan, Lu Yuan, Lei Zhang, Zhangyang Wang

Chasing Sparsity in Vision Transformers: An End-to-End Exploration Codes for [Preprint] Chasing Sparsity in Vision Transformers: An End-to-End Explora

VITA 64 Dec 8, 2022
[CVPR 2022] "The Principle of Diversity: Training Stronger Vision Transformers Calls for Reducing All Levels of Redundancy" by Tianlong Chen, Zhenyu Zhang, Yu Cheng, Ahmed Awadallah, Zhangyang Wang

The Principle of Diversity: Training Stronger Vision Transformers Calls for Reducing All Levels of Redundancy Codes for this paper: [CVPR 2022] The Pr

VITA 16 Nov 26, 2022
[ICCV'2021] "SSH: A Self-Supervised Framework for Image Harmonization", Yifan Jiang, He Zhang, Jianming Zhang, Yilin Wang, Zhe Lin, Kalyan Sunkavalli, Simon Chen, Sohrab Amirghodsi, Sarah Kong, Zhangyang Wang

SSH: A Self-Supervised Framework for Image Harmonization (ICCV 2021) code for SSH Representative Examples Main Pipeline RealHM DataSet Google Drive Pr

VITA 86 Dec 2, 2022
[ICLR 2021] "Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective" by Wuyang Chen, Xinyu Gong, Zhangyang Wang

Neural Architecture Search on ImageNet in Four GPU Hours: A Theoretically Inspired Perspective [PDF] Wuyang Chen, Xinyu Gong, Zhangyang Wang In ICLR 2

VITA 156 Nov 28, 2022
"SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image", Dejia Xu, Yifan Jiang, Peihao Wang, Zhiwen Fan, Humphrey Shi, Zhangyang Wang

SinNeRF: Training Neural Radiance Fields on Complex Scenes from a Single Image [Paper] [Website] Pipeline Code Environment pip install -r requirements

VITA 250 Jan 5, 2023
This is the official PyTorch implementation of the paper "TransFG: A Transformer Architecture for Fine-grained Recognition" (Ju He, Jie-Neng Chen, Shuai Liu, Adam Kortylewski, Cheng Yang, Yutong Bai, Changhu Wang, Alan Yuille).

TransFG: A Transformer Architecture for Fine-grained Recognition Official PyTorch code for the paper: TransFG: A Transformer Architecture for Fine-gra

Ju He 307 Jan 3, 2023
Implementation of "Efficient Regional Memory Network for Video Object Segmentation" (Xie et al., CVPR 2021).

RMNet This repository contains the source code for the paper Efficient Regional Memory Network for Video Object Segmentation. Cite this work @inprocee

Haozhe Xie 76 Dec 14, 2022
[ICLR 2021, Spotlight] Large Scale Image Completion via Co-Modulated Generative Adversarial Networks

Large Scale Image Completion via Co-Modulated Generative Adversarial Networks, ICLR 2021 (Spotlight) Demo | Paper [NEW!] Time to play with our interac

Shengyu Zhao 373 Jan 2, 2023
Code for "The Intrinsic Dimension of Images and Its Impact on Learning" - ICLR 2021 Spotlight

dimensions Estimating the instrinsic dimensionality of image datasets Code for: The Intrinsic Dimensionaity of Images and Its Impact On Learning - Phi

Phil Pope 41 Dec 10, 2022
Official Implementation of 'UPDeT: Universal Multi-agent Reinforcement Learning via Policy Decoupling with Transformers' ICLR 2021(spotlight)

UPDeT Official Implementation of UPDeT: Universal Multi-agent Reinforcement Learning via Policy Decoupling with Transformers (ICLR 2021 spotlight) The

hhhusiyi 96 Dec 22, 2022
Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy" (ICLR 2022 Spotlight)

About Code release for Anomaly Transformer: Time Series Anomaly Detection with Association Discrepancy (ICLR 2022 Spotlight)

THUML @ Tsinghua University 221 Dec 31, 2022
Based on the paper "Geometry-aware Instance-reweighted Adversarial Training" ICLR 2021 oral

Geometry-aware Instance-reweighted Adversarial Training This repository provides codes for Geometry-aware Instance-reweighted Adversarial Training (ht

Jingfeng 47 Dec 22, 2022
PyTorch implementation for Score-Based Generative Modeling through Stochastic Differential Equations (ICLR 2021, Oral)

Score-Based Generative Modeling through Stochastic Differential Equations This repo contains a PyTorch implementation for the paper Score-Based Genera

Yang Song 757 Jan 4, 2023
Code for the ICML 2021 paper "Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Training and Effective Adaptation", Haoxiang Wang, Han Zhao, Bo Li.

Bridging Multi-Task Learning and Meta-Learning Code for the ICML 2021 paper "Bridging Multi-Task Learning and Meta-Learning: Towards Efficient Trainin

AI Secure 57 Dec 15, 2022
Official codebase for ICLR oral paper Unsupervised Vision-Language Grammar Induction with Shared Structure Modeling

CLIORA This is the official codebase for ICLR oral paper: Unsupervised Vision-Language Grammar Induction with Shared Structure Modeling. We introduce

Bo Wan 32 Dec 23, 2022