Fast Differentiable Matrix Sqrt Root

Last update: Dec 30, 2022

Overview

Fast Differentiable Matrix Sqrt Root

Geometric Interpretation of Matrix Square Root and Inverse Square Root

This repository constains the official Pytorch implementation of ICLR 22 paper "Fast Differentiable Matrix Square Root".

Usages

Check torch_utils.py for the implementation. Minimal exemplery usage is given as follows:

# Import and define function
from torch_utils import *
FastMatSqrt=MPA_Lya.apply
FastInvSqrt=MPA_Lya_Inv.apply

# For any batched matrices, compute their square root or inverse square root:
rand_matrix = torch.randn(5,32,32)
rand_cov = rand_matrix.bmm(rand_matrix.transpose(1,2))
rand_cov_sqrt = FastMatSqrt(rand_cov)
rand_inv_sqrt = FastInvSqrt(rand_cov)

Computer Vision Experiments

All the codes for the following experiments are available:

Citation

Please consider citing our paper if you think the code is helpful to your research.

@inproceedings{song2022fast,
  title={Fast Differentiable Matrix Square Root},
  author={Song, Yue and Sebe, Nicu and Wang, Wei},
  booktitle={ICLR},
  year={2022}
}

Contact

If you have any questions or suggestions, please feel free to contact me

[email protected]

UV matrix decompostion using movielens dataset

UV-matrix-decompostion-with-kfold UV matrix decompostion using movielens dataset upload the 'ratings.dat' file install the following python libraries

2 Oct 18, 2022

TART - A PyTorch implementation for Transition Matrix Representation of Trees with Transposed Convolutions

TART This project is a PyTorch implementation for Transition Matrix Representati

2 Jan 19, 2022

A script written in Python that returns a consensus string and profile matrix of a given DNA string(s) in FASTA format.

1 Feb 1, 2022

Applications using the GTN library and code to reproduce experiments in "Differentiable Weighted Finite-State Transducers"

gtn_applications An applications library using GTN. Current examples include: Offline handwriting recognition Automatic speech recognition Installing

68 Dec 29, 2022

Performant, differentiable reinforcement learning

deluca Performant, differentiable reinforcement learning Notes This is pre-alpha software and is undergoing a number of core changes. Updates to follo

114 Dec 27, 2022

Differentiable molecular simulation of proteins with a coarse-grained potential

Differentiable molecular simulation of proteins with a coarse-grained potential This repository contains the learned potential, simulation scripts and

44 Dec 10, 2022

UnsupervisedR&R: Unsupervised Pointcloud Registration via Differentiable Rendering

UnsupervisedR&R: Unsupervised Pointcloud Registration via Differentiable Rendering This repository holds all the code and data for our recent work on

118 Dec 6, 2022

Code for the ECCV2020 paper "A Differentiable Recurrent Surface for Asynchronous Event-Based Data"

A Differentiable Recurrent Surface for Asynchronous Event-Based Data Code for the ECCV2020 paper "A Differentiable Recurrent Surface for Asynchronous

21 Oct 5, 2022

Differentiable Optimizers with Perturbations in Pytorch

Differentiable Optimizers with Perturbations in PyTorch This contains a PyTorch implementation of Differentiable Optimizers with Perturbations in Tens

54 Jun 22, 2022

Comments

A question of compute P and Q weithou P = I - P, Q = I - Q

(https://github.com/KingJamesSong/FastDifferentiableMatSqrt/blob/466c9bf5f7ca5c3a405e104197e63f20b1c70114/torch_utils.py#L28-L40)

why not $p \underline{} sqrt = \mathbf{I} - p \underline{} sqrt,q \underline{} sqrt=\mathbf{I} - q \underline{} sqrt $ This seems inconsistent with the formula in the paper

opened by HongLouyemeng 4
Inverse of singular matrix doesn't exist, so what does the solution represent ??

Hi,

I was using the code for the inversion of a covariance matrix, but the covariance matrix I have created is from a single data point i.e [email protected] where A is a vector of shape (n,1).

So [email protected] is a rank1 matrix of shape (n,n). So the inverse should not exist for this matrix. But, when calculating the inverse of this matrix, the method provides a solution.

So I wanted to ask what is this solution. Since the method is iterative, so it approximates the result, but to what result is it approximating ??

opened by lavish619 2

Fast Differentiable Matrix Sqrt Root

Related tags

Overview

Fast Differentiable Matrix Sqrt Root

Usages

Computer Vision Experiments

Citation

Contact

You might also like...

UV matrix decompostion using movielens dataset

TART - A PyTorch implementation for Transition Matrix Representation of Trees with Transposed Convolutions

A script written in Python that returns a consensus string and profile matrix of a given DNA string(s) in FASTA format.

Applications using the GTN library and code to reproduce experiments in "Differentiable Weighted Finite-State Transducers"

Performant, differentiable reinforcement learning

Differentiable molecular simulation of proteins with a coarse-grained potential

UnsupervisedR&R: Unsupervised Pointcloud Registration via Differentiable Rendering

Code for the ECCV2020 paper "A Differentiable Recurrent Surface for Asynchronous Event-Based Data"

Differentiable Optimizers with Perturbations in Pytorch

Comments

A question of compute P and Q weithou P = I - P, Q = I - Q

Inverse of singular matrix doesn't exist, so what does the solution represent ??

Owner

YueSong

Implementation of fast algorithms for Maximum Spanning Tree (MST) parsing that includes fast ArcMax+Reweighting+Tarjan algorithm for single-root dependency parsing.

Differentiable Neural Computers, Sparse Access Memory and Sparse Differentiable Neural Computers, for Pytorch

Square Root Bundle Adjustment for Large-Scale Reconstruction

A simple root calculater for python

FirmWire is a full-system baseband firmware emulation platform for fuzzing, debugging, and root-cause analysis of smartphone baseband firmwares

Fast, differentiable sorting and ranking in PyTorch

[ICLR 2021] Is Attention Better Than Matrix Decomposition?

Graph-based community clustering approach to extract protein domains from a predicted aligned error matrix

A numpy-based implementation of RANSAC for fundamental matrix and homography estimation. The degeneracy updating and local optimization components are included and optional.

Implementation of SSMF: Shifting Seasonal Matrix Factorization