Pytorch Implementation of paper "Noisy Natural Gradient as Variational Inference"

Overview

Noisy Natural Gradient as Variational Inference

PyTorch implementation of Noisy Natural Gradient as Variational Inference.

Requirements

  • Python 3
  • Pytorch
  • visdom

Comments

  • This paper is about how to optimize bayesian neural network which has matrix variate gaussian distribution.
  • This implementation contains Noisy Adam optimizer which is for Fully Factorized Gaussian(FFG) distribution, and Noisy KFAC optimizer which is for Matrix Variate Gaussian(MVG) distribution.
  • These optimizers only work with bayesian network which has specific structure that I will mention below.
  • Currently only linear layer is available.

Experimental comments

  • I addded a lr scheduler to noisy KFAC because loss is exploded during training. I guess this happens because of slight approximation.
  • For MNIST training noisy KFAC is 15-20x slower than noisy Adam, as mentioned in paper.
  • I guess the noisy KFAC needs more epochs to train simple neural network structure like 2 linear layers.

Usage

Currently only MNIST dataset are currently supported, and only fully connected layer is implemented.

Options

  • model : Fully Factorized Gaussian(FFG) or Matrix Variate Gaussian(MVG)
  • n : total train dataset size. need this value for optimizer.
  • eps : parameter for optimizer. Default to 1e-8.
  • initial_size : initial input tensor size. Default to 784, size of MNIST data.
  • label_size : label size. Default to 10, size of MNIST label.

More details in option_parser.py

Train

$ python train.py --model=FFG --batch_size=100 --lr=1e-3 --dataset=MNIST
$ python train.py --model=MVG --batch_size=100 --lr=1e-2 --dataset=MNIST --n=60000

Visualize

  • To visualize intermediate results and loss plots, run python -m visdom.server and go to the URL http://localhost:8097

Test

$ python test.py --epoch=20

Training Graphs

1. MNIST

  • network is consist of 2 linear layers.
  • FFG optimized by noisy Adam : epoch 20, lr 1e-3

  • MVG optimized by noisy KFAC : epoch 100, lr 1e-2, decay 0.1 for every 30 epochs
  • Need to tune learning rate.

Implementation detail

  • Optimizing parameter procedure is consists of 2 steps, Calculating gradient and Applying to bayeisan parameters.
  • Before forward, network samples parameters with means & variances.
  • Usually calling step function updates parameters, but not this case. After calling step function, you have to update bayesian parameters. Look at the ffg_model.py

TODOs

  • More benchmark cases
  • Supports bayesian convolution
  • Implement Block Tridiagonal Covariance, which is dependent between layers.

Code reference

Visualization code(visualizer.py, utils.py) references to pytorch-CycleGAN-and-pix2pix(https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix) by Jun-Yan Zhu

Author

Tony Kim

You might also like...
An implementation of Geoffrey Hinton's paper
An implementation of Geoffrey Hinton's paper "How to represent part-whole hierarchies in a neural network" in Pytorch.

GLOM An implementation of Geoffrey Hinton's paper "How to represent part-whole hierarchies in a neural network" for MNIST Dataset. To understand this

Official implementation of our CVPR2021 paper
Official implementation of our CVPR2021 paper "OTA: Optimal Transport Assignment for Object Detection" in Pytorch.

OTA: Optimal Transport Assignment for Object Detection This project provides an implementation for our CVPR2021 paper "OTA: Optimal Transport Assignme

This is the official PyTorch implementation of the paper
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

[PyTorch] Official implementation of CVPR2021 paper
[PyTorch] Official implementation of CVPR2021 paper "PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency". https://arxiv.org/abs/2103.05465

PointDSC repository PyTorch implementation of PointDSC for CVPR'2021 paper "PointDSC: Robust Point Cloud Registration using Deep Spatial Consistency",

PyTorch implementation of the Deep SLDA method from our CVPRW-2020 paper
PyTorch implementation of the Deep SLDA method from our CVPRW-2020 paper "Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis"

Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis This is a PyTorch implementation of the Deep Streaming Linear Discriminant

The official pytorch implementation of our paper "Is Space-Time Attention All You Need for Video Understanding?"

TimeSformer This is an official pytorch implementation of Is Space-Time Attention All You Need for Video Understanding?. In this repository, we provid

PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement.
PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement.

DECOR-GAN PyTorch 1.5 implementation for paper DECOR-GAN: 3D Shape Detailization by Conditional Refinement, Zhiqin Chen, Vladimir G. Kim, Matthew Fish

Pytorch implementation of our paper under review — Lottery Jackpots Exist in Pre-trained Models

Lottery Jackpots Exist in Pre-trained Models (Paper Link) Requirements Python = 3.7.4 Pytorch = 1.6.1 Torchvision = 0.4.1 Reproduce the Experiment

Official Pytorch Implementation of:
Official Pytorch Implementation of: "ImageNet-21K Pretraining for the Masses"(2021) paper

ImageNet-21K Pretraining for the Masses Paper | Pretrained models Official PyTorch Implementation Tal Ridnik, Emanuel Ben-Baruch, Asaf Noy, Lihi Zelni

Owner
Tony JiHyun Kim
CEO/Tech Lead @PostAlpine Co., Ltd.
Tony JiHyun Kim
HashNeRF-pytorch - Pure PyTorch Implementation of NVIDIA paper on Instant Training of Neural Graphics primitives

HashNeRF-pytorch Instant-NGP recently introduced a Multi-resolution Hash Encodin

Yash Sanjay Bhalgat 616 Jan 6, 2023
ALBERT-pytorch-implementation - ALBERT pytorch implementation

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

BG Kim 3 Oct 6, 2022
The LaTeX and Python code for generating the paper, experiments' results and visualizations reported in each paper is available (whenever possible) in the paper's directory

This repository contains the software implementation of most algorithms used or developed in my research. The LaTeX and Python code for generating the

João Fonseca 3 Jan 3, 2023
Official PyTorch implementation for paper Context Matters: Graph-based Self-supervised Representation Learning for Medical Images

Context Matters: Graph-based Self-supervised Representation Learning for Medical Images Official PyTorch implementation for paper Context Matters: Gra

null 49 Nov 23, 2022
A PyTorch re-implementation of the paper 'Exploring Simple Siamese Representation Learning'. Reproduced the 67.8% Top1 Acc on ImageNet.

Exploring simple siamese representation learning This is a PyTorch re-implementation of the SimSiam paper on ImageNet dataset. The results match that

Taojiannan Yang 72 Nov 9, 2022
This is a Pytorch implementation of the paper: Self-Supervised Graph Transformer on Large-Scale Molecular Data.

This is a Pytorch implementation of the paper: Self-Supervised Graph Transformer on Large-Scale Molecular Data.

null 212 Dec 25, 2022
Official pytorch implementation of paper "Image-to-image Translation via Hierarchical Style Disentanglement".

HiSD: Image-to-image Translation via Hierarchical Style Disentanglement Official pytorch implementation of paper "Image-to-image Translation

null 364 Dec 14, 2022
PyTorch implementation of paper "Neural Scene Flow Fields for Space-Time View Synthesis of Dynamic Scenes", CVPR 2021

Neural Scene Flow Fields PyTorch implementation of paper "Neural Scene Flow Fields for Space-Time View Synthesis of Dynamic Scenes", CVPR 20

Zhengqi Li 585 Jan 4, 2023
Official pytorch implementation of paper "Inception Convolution with Efficient Dilation Search" (CVPR 2021 Oral).

IC-Conv This repository is an official implementation of the paper Inception Convolution with Efficient Dilation Search. Getting Started Download Imag

Jie Liu 111 Dec 31, 2022
Official implementation of our paper "LLA: Loss-aware Label Assignment for Dense Pedestrian Detection" in Pytorch.

LLA: Loss-aware Label Assignment for Dense Pedestrian Detection This project provides an implementation for "LLA: Loss-aware Label Assignment for Dens

null 35 Dec 6, 2022