PyTorch implementation of CVPR'18 - Perturbative Neural Networks

Overview

Perturbative Neural Networks (PNN)

This is an attempt to reproduce results in Perturbative Neural Networks paper. See original repo for details.

Motivation

The original implementation used regular convolutions in the first layer, and the remaining layers used fanout of 1, which means each input channel was perturbed with a single noise mask.

However, the biggest issue with the original implementation is that test accuracy was calculated incorrectly. Instead of the usual method of calculating ratio of correct samples to total samples in the test dataset, the authors calculated accuracy on per batch basis, and applied smoothing weight (test_accuracy = 0.7 * prev_batch_accuracy + 0.3 * current_batch_accuracy).

Here's how this method (reported) compares to the proper accuracy calculation (actual):

img

For this model run (noiseresnet18 on CIFAR10), the code in original repo would report best test accuracy 90.53%, while the actual best test accuracy is 85.91%

After correcting this issue, I ran large number of experiments trying to see if perturbing input with noise masks would provide any benefit, and my conclusion is that it does not.

Here's for example, the difference between ResNet18-like models: a baseline model with reduced number of filters to keep the same parameter count, a model where all layers except first one use only 1x1 convolutions (no noise), and a model where all layers except first one use perturbations followed by 1x1 convolutions. All three models have ~5.5M parameters:

img

The accuracy difference between regular resnet baseline and PNN remains ~5% throughout the training, and the addition of noise masks results in less than 1% improvement over equivalently "crippled" resnet without any noise applied.

Implementation details

Most of the modifications are contained in the PerturbLayer class. Here are the main changes from the original code:

--first_filter_size and --filter_size arguments control the type of the first layer, and the remaining layers, correspondingly. A value of 0 turns the layer into a perturbation layer, as described in the paper. Any value n > 0 will turn the layer into a regular convolutional layer with filter size n. The original implementation only supports first_filter_size=7, and filter_size=0.

--nmasks specifies number of noise masks to apply to each input channel. This is "fanout" parameter mentioned in the paper. The original implementation only supports nmasks=1.

--unique_masks specifies whether to use different sets of nmasks noise masks for each input channel. --no-unique_masks forces the same set of nmasks to be used for all input channels.

--train_masks enables treating noise masks as regular parameters, and optimizes their values during training at the same time as model weights.

--mix_maps adds second 1x1 convolutional layer after perturbed input channels are combined with the first 1x1 convolution. Without this second 1x1 "mixing" layer, there is no information exchange between input channels all the way until the softmax layer in the end. Note that it's not needed when --nmasks is 1, because then the first 1x1 convolutional layer already plays this role.

Other arguments allow changing noise type (uniform or normal), pooling type (max or avg), activation function (relu, rrelu, prelu, elu, selu, tanh, sigmoid), whether to apply activation function in the first layer (--use_act, immediately after perturbing the input RGB channels, this results in some information loss), whether to scale noise level in the first layer, and --debug argument prints out values of input, noise, and output for every update step to verify that noise is being applied correctly.

Three different models are supported: perturb_resnet18, cifarnet (6 conv layers, followed by a fully connected layer), and lenet (3 conv. layers followed by a fully connected layer). In addition, I included the baseline ResNet-18 model resnet18 taken from here, and noiseresnet18 model from the original repo. Note that perturb_resnet18 model is flexible enough to replace both baseline and noiseresnet18 models, using appropriate arguments.

Results

CIFAR-10:

  1. Baseline (regular ResNet18 with 3x3 convolutions, number of filters reduced to match PNN parameter count) Test Accuracy: 91.8%
python main.py --net-type 'resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 44 --batch-size 10 --learning-rate 1e-3
  1. Original implementation (equivalent to running the code from the original repo). Test Accuracy: 85.7%
python main.py --net-type 'noiseresnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 10 --learning-rate 1e-4 --first_filter_size 7
  1. Same as above, but changing first_filter_size argument to 3 improves the accuracy to 86.2%

  2. Same as above, but without any noise (resnet18 with 3x3 convolutions in the first layer, and 1x1 in remaining layers). Test Accuracy: 85.5%

python main.py --net-type 'perturb_resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 16 --learning-rate 1e-3 --first_filter_size 3 --filter_size 1 
  1. PNN with all uniform noise in all layers (including the first layer). Test Accuracy: 72.6%
python main.py --net-type 'perturb_resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 16 --learning-rate 1e-3 --first_filter_size 0 --filter_size 0 --nmasks 1 
  1. PNN with noise masks in all layers except the first layer, which uses regular 3x3 convolutions with fanout=64. Internally fanout is implemented with grouped 1x1 convolutions. Note: --unique_masks arg creates unique set of masks for each input channel, in every layer, and --mix_maps argument which uses extra 1x1 convolutional layer in all perturbation layers. Test Accuracy: 82.7%
python main.py --net-type 'perturb_resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 16 --learning-rate 1e-3 --first_filter_size 3 --filter_size 0 --nmasks 64 --unique_masks --mix_maps
  1. Same as above, but with --no-unique_masks argument, which means that the same set of masks is used for each input channel. Test Accuracy: 82.4%
python main.py --net-type 'perturb_resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 16 --learning-rate 1e-3 --first_filter_size 3 --filter_size 0 --nmasks 64 --no-unique_masks

Experiments 6 and 7 are the closest to what was described in the paper.

  1. training the noise masks (updated each batch, at the same time as regular model parameters). Test Accuracy: 85.9%

python main.py --net-type 'perturb_resnet18' --dataset-test 'CIFAR10' --dataset-train 'CIFAR10' --nfilters 128 --batch-size 16 --learning-rate 1e-3 --first_filter_size 3 --filter_size 0 --nmasks 64 --no-unique_masks --train_masks

Weakness of reasoning:

Section 3.3: "given the known input x and convolution transformation matrix A, we can always solve for the matching noise perturbation matrix N".

While for any given single input sample PNN might be able to find the weights required to match the output of a CNN, it does not follow that it can find weights to do that for all input samples in the dataset.

Section 3.4: The result of a single convolution operation is represented as a value of the center pixel Xc in a patch X, plus some quantity Nc (a function of filter weights W and the neighboring pixels of Xc): Y = XW = Xc + Nc. The claim is: "Establishing that Nc behaves like additive perturbation noise, will allows us to relate the CNN formulation to the PNN formulation".

Even if Nc statistically behaves like random noise does not mean it can be replaced with random noise. The random noise in PNN does not depend on values of neighboring pixels in the patch, unlike Nc in a regular convolution. PNN layer lacks the main feature extraction property of a regular convolution: it cannot directly match any spatial patterns with a filter.

Conclusion

It appears that perturbing layer inputs with noise does not provide any significant benefit. Simple 1x1 convolutions without noise masks provide similar performance. No matter how we apply noise masks, the accuracy drop resulting from using 1x1 filters is severe (~5% on CIFAR-10 even when not modifying the first layer). The results published by the authors are invalid due to incorrect accuracy calculation method.

Owner
Michael Klachko
Michael Klachko
基于pytorch构建cyclegan示例

cyclegan-demo 基于Pytorch构建CycleGAN示例 如何运行 准备数据集 将数据集整理成4个文件,分别命名为 trainA, trainB:训练集,A、B代表两类图片 testA, testB:测试集,A、B代表两类图片 例如 D:\CODE\CYCLEGAN-DEMO\DATA

Koorye 3 Oct 18, 2022
level1-image-classification-level1-recsys-09 created by GitHub Classroom

level1-image-classification-level1-recsys-09 ❗ 주제 설명 COVID-19 Pandemic 상황 속 마스크 착용 유무 판단 시스템 구축 마스크 착용 여부, 성별, 나이 총 세가지 기준에 따라 총 18개의 class로 구분하는 모델 ?

6 Mar 17, 2022
Experimental code for paper: Generative Adversarial Networks as Variational Training of Energy Based Models

Experimental code for paper: Generative Adversarial Networks as Variational Training of Energy Based Models, under review at ICLR 2017 requirements: T

Shuangfei Zhai 18 Mar 05, 2022
Statistical and Algorithmic Investing Strategies for Everyone

Eiten - Algorithmic Investing Strategies for Everyone Eiten is an open source toolkit by Tradytics that implements various statistical and algorithmic

Tradytics 2.5k Jan 02, 2023
Pytorch implementation of "Forward Thinking: Building and Training Neural Networks One Layer at a Time"

forward-thinking-pytorch Pytorch implementation of Forward Thinking: Building and Training Neural Networks One Layer at a Time Requirements Python 2.7

Kim Heecheol 65 Oct 06, 2022
LAnguage Model Analysis

LAMA: LAnguage Model Analysis LAMA is a probe for analyzing the factual and commonsense knowledge contained in pretrained language models. The dataset

Meta Research 960 Jan 08, 2023
Learning the Beauty in Songs: Neural Singing Voice Beautifier; ACL 2022 (Main conference); Official code

Learning the Beauty in Songs: Neural Singing Voice Beautifier Jinglin Liu, Chengxi Li, Yi Ren, Zhiying Zhu, Zhou Zhao Zhejiang University ACL 2022 Mai

Jinglin Liu 257 Dec 30, 2022
GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration

GenGNN: A Generic FPGA Framework for Graph Neural Network Acceleration Stefan Abi-Karam*, Yuqi He*, Rishov Sarkar*, Lakshmi Sathidevi, Zihang Qiao, Co

Sharc-Lab 19 Dec 15, 2022
Emotion Recognition from Facial Images

Reconhecimento de Emoções a partir de imagens faciais Este projeto implementa um classificador simples que utiliza técncias de deep learning e transfe

Gabriel 2 Feb 09, 2022
RATCHET is a Medical Transformer for Chest X-ray Diagnosis and Reporting

RATCHET: RAdiological Text Captioning for Human Examined Thoraxes RATCHET is a Medical Transformer for Chest X-ray Diagnosis and Reporting. Based on t

26 Nov 14, 2022
ALBERT-pytorch-implementation - ALBERT pytorch implementation

ALBERT-pytorch-implementation developing... 모델의 개념이해를 돕기 위한 구현물로 현재 변수명을 상세히 적었고

BG Kim 3 Oct 06, 2022
LaneAF: Robust Multi-Lane Detection with Affinity Fields

LaneAF: Robust Multi-Lane Detection with Affinity Fields This repository contains Pytorch code for training and testing LaneAF lane detection models i

155 Dec 17, 2022
FSL-Mate: A collection of resources for few-shot learning (FSL).

FSL-Mate is a collection of resources for few-shot learning (FSL). In particular, FSL-Mate currently contains FewShotPapers: a paper list which tracks

Yaqing Wang 1.5k Jan 08, 2023
💛 Code and Dataset for our EMNLP 2021 paper: "Perspective-taking and Pragmatics for Generating Empathetic Responses Focused on Emotion Causes"

Perspective-taking and Pragmatics for Generating Empathetic Responses Focused on Emotion Causes Official PyTorch implementation and EmoCause evaluatio

Hyunwoo Kim 51 Jan 06, 2023
covid question answering datasets and fine tuned models

Covid-QA Fine tuned models for question answering on Covid-19 data. Hosted Inference This model has been contributed to huggingface.Click here to see

Abhijith Neil Abraham 19 Sep 09, 2021
Repository of Jupyter notebook tutorials for teaching the Deep Learning Course at the University of Amsterdam (MSc AI), Fall 2020

Repository of Jupyter notebook tutorials for teaching the Deep Learning Course at the University of Amsterdam (MSc AI), Fall 2020

Phillip Lippe 1.1k Jan 07, 2023
Neural Motion Learner With Python

Neural Motion Learner Introduction This work is to extract skeletal structure from volumetric observations and to learn motion dynamics from the detec

Jinseok Bae 14 Nov 28, 2022
Implementation for our ICCV 2021 paper: Dual-Camera Super-Resolution with Aligned Attention Modules

DCSR: Dual Camera Super-Resolution Implementation for our ICCV 2021 oral paper: Dual-Camera Super-Resolution with Aligned Attention Modules paper | pr

Tengfei Wang 110 Dec 20, 2022
Robustness between the worst and average case

Robustness between the worst and average case A repository that implements intermediate robustness training and evaluation from the NeurIPS 2021 paper

CMU Locus Lab 16 Dec 02, 2022