A Python package for generating concise, high-quality summaries of a probability distribution

Overview

GoodPoints

A Python package for generating concise, high-quality summaries of a probability distribution

GoodPoints is a collection of tools for compressing a distribution more effectively than independent sampling:

  • Given an initial summary of n input points, kernel thinning returns s << n output points with comparable integration error across a reproducing kernel Hilbert space
  • Compress++ reduces the runtime of generic thinning algorithms with minimal loss in accuracy

Installation

To install the goodpoints package, use the following pip command:

pip install goodpoints

Getting started

The primary kernel thinning function is thin in the kt module:

from goodpoints import kt
coreset = kt.thin(X, m, split_kernel, swap_kernel, delta=0.5, seed=123, store_K=False)
    """Returns kernel thinning coreset of size floor(n/2^m) as row indices into X
    
    Args:
      X: Input sequence of sample points with shape (n, d)
      m: Number of halving rounds
      split_kernel: Kernel function used by KT-SPLIT (typically a square-root kernel, krt);
        split_kernel(y,X) returns array of kernel evaluations between y and each row of X
      swap_kernel: Kernel function used by KT-SWAP (typically the target kernel, k);
        swap_kernel(y,X) returns array of kernel evaluations between y and each row of X
      delta: Run KT-SPLIT with constant failure probabilities delta_i = delta/n
      seed: Random seed to set prior to generation; if None, no seed will be set
      store_K: If False, runs O(nd) space version which does not store kernel
        matrix; if True, stores n x n kernel matrix
    """

For example uses, please refer to the notebook examples/kt/run_kt_experiment.ipynb.

The primary Compress++ function is compresspp in the compress module:

from goodpoints import compress
coreset = compress.compresspp(X, halve, thin, g)
    """Returns Compress++(g) coreset of size sqrt(n) as row indices into X

    Args: 
        X: Input sequence of sample points with shape (n, d)
        halve: Function that takes in an (n', d) numpy array Y and returns 
          floor(n'/2) distinct row indices into Y, identifying a halved coreset
        thin: Function that takes in an (n', d) numpy array Y and returns
          2^g sqrt(n') row indices into Y, identifying a thinned coreset
        g: Oversampling factor
    """

For example uses, please refer to the code examples/compress/construct_compresspp_coresets.py.

Examples

Code in the examples directory uses the goodpoints package to recreate the experiments of the following research papers.


Kernel Thinning

@article{dwivedi2021kernel,
  title={Kernel Thinning},
  author={Raaz Dwivedi and Lester Mackey},
  journal={arXiv preprint arXiv:2105.05842},
  year={2021}
}
  1. The script examples/kt/submit_jobs_run_kt.py reproduces the vignette experiments of Kernel Thinning on a Slurm cluster by executing examples/kt/run_kt_experiment.ipynb with appropriate parameters. For the MCMC examples, it assumes that necessary data was downloaded and pre-processed following the steps listed in examples/kt/preprocess_mcmc_data.ipynb, where in the last code block we report the median heuristic based bandwidth parameteters (along with the code to compute it).
  2. After all results have been generated, the notebook plot_results.ipynb can be used to reproduce the figures of Kernel Thinning.

Generalized Kernel Thinning

@article{dwivedi2021generalized,
  title={Generalized Kernel Thinning},
  author={Raaz Dwivedi and Lester Mackey},
  journal={arXiv preprint arXiv:2110.01593},
  year={2021}
}
  1. The script examples/gkt/submit_gkt_jobs.py reproduces the vignette experiments of Generalized Kernel Thinning on a Slurm cluster by executing examples/gkt/run_generalized_kt_experiment.ipynb with appropriate parameters. For the MCMC examples, it assumes that necessary data was downloaded and pre-processed following the steps listed in examples/kt/preprocess_mcmc_data.ipynb.
  2. Once the coresets are generated, examples/gkt/compute_test_function_errors.ipynb can be used to generate integration errors for different test functions.
  3. After all results have been generated, the notebook examples/gkt/plot_gkt_results.ipynb can be used to reproduce the figures of Generalized Kernel Thinning.

Distribution Compression in Near-linear Time

@article{shetti2021distribution,
  title={Distribution Compression in Near-linear Time},
  author={Abhishek Shetty and Raaz Dwivedi and Lester Mackey},
  journal={arXiv preprint arXiv:2111.07941},
  year={2021}
}
  1. The notebook examples/compress/script_to_deploy_jobs.ipynb reproduces the experiments of Distribution Compression in Near-linear Time in the following manner: 1a. It generates various coresets and computes their mmds by executing examples/compress/construct_{THIN}_coresets.py for THIN in {compresspp, kt, st, herding} with appropriate parameters, where the flag kt stands for kernel thinning, st stands for standard thinning (choosing every t-th point), and herding refers to kernel herding. 1b. It compute the runtimes of different algorithms by executing examples/compress/run_time.py. 1c. For the MCMC examples, it assumes that necessary data was downloaded and pre-processed following the steps listed in examples/kt/preprocess_mcmc_data.ipynb. 1d. The notebook currently deploys these jobs on a slurm cluster, but setting deploy_slurm = False in examples/compress/script_to_deploy_jobs.ipynb will submit the jobs as independent python calls on terminal.
  2. After all results have been generated, the notebook examples/compress/plot_compress_results.ipynb can be used to reproduce the figures of Distribution Compression in Near-linear Time.
  3. The script examples/compress/construct_compresspp_coresets.py contains the function recursive_halving that converts a halving algorithm into a thinning algorithm by recursively halving.
  4. The script examples/compress/construct_herding_coresets.py contains the herding function that runs kernel herding algorithm introduced by Yutian Chen, Max Welling, and Alex Smola.

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.

Owner
Microsoft
Open source projects and samples from Microsoft
Microsoft
Repo for Photon-Starved Scene Inference using Single Photon Cameras, ICCV 2021

Photon-Starved Scene Inference using Single Photon Cameras ICCV 2021 Arxiv Project Video Bhavya Goyal, Mohit Gupta University of Wisconsin-Madison Abs

Bhavya Goyal 5 Nov 15, 2022
Implementation for Paper "Inverting Generative Adversarial Renderer for Face Reconstruction"

StyleGAR TODO: add arxiv link Implementation of Inverting Generative Adversarial Renderer for Face Reconstruction TODO: for test Currently, some model

155 Oct 27, 2022
Your interactive network visualizing dashboard

Your interactive network visualizing dashboard Documentation: Here What is Jaal Jaal is a python based interactive network visualizing tool built usin

Mohit 177 Jan 04, 2023
Code for Blind Image Decomposition (BID) and Blind Image Decomposition network (BIDeN).

arXiv, porject page, paper Blind Image Decomposition (BID) Blind Image Decomposition is a novel task. The task requires separating a superimposed imag

64 Dec 20, 2022
RoMA: Robust Model Adaptation for Offline Model-based Optimization

RoMA: Robust Model Adaptation for Offline Model-based Optimization Implementation of RoMA: Robust Model Adaptation for Offline Model-based Optimizatio

9 Oct 31, 2022
Piotr - IoT firmware emulation instrumentation for training and research

Piotr: Pythonic IoT exploitation and Research Introduction to Piotr Piotr is an emulation helper for Qemu that provides a convenient way to create, sh

Damien Cauquil 51 Nov 09, 2022
Nsdf: A mesh SDF with just some code we can directly paste into our raymarcher

nsdf Representing SDFs of arbitrary meshes has been a bit tricky so far. Express

Jan Ivanecky 5 Feb 18, 2022
Sequence lineage information extracted from RKI sequence data repo

Pango lineage information for German SARS-CoV-2 sequences This repository contains a join of the metadata and pango lineage tables of all German SARS-

Cornelius Roemer 24 Oct 26, 2022
PyTorch implementation of DUL (Data Uncertainty Learning in Face Recognition, CVPR2020)

PyTorch implementation of DUL (Data Uncertainty Learning in Face Recognition, CVPR2020)

Mouxiao Huang 20 Nov 15, 2022
Campsite Reservation Finder

yellowstone-camping UPDATE: yellowstone-camping is being expanded and renamed to camply. The updated tool now interfaces with the Recreation.gov API a

Justin Flannery 233 Jan 08, 2023
Radar-to-Lidar: Heterogeneous Place Recognition via Joint Learning

radar-to-lidar-place-recognition This page is the coder of a pre-print, implemented by PyTorch. If you have some questions on this project, please fee

Huan Yin 37 Oct 09, 2022
Implementation of Pooling by Sliced-Wasserstein Embedding (NeurIPS 2021)

PSWE: Pooling by Sliced-Wasserstein Embedding (NeurIPS 2021) PSWE is a permutation-invariant feature aggregation/pooling method based on sliced-Wasser

Navid Naderializadeh 3 May 06, 2022
Pytorch implementation of PCT: Point Cloud Transformer

PCT: Point Cloud Transformer This is a Pytorch implementation of PCT: Point Cloud Transformer.

Yi_Zhang 265 Dec 22, 2022
Material for my PyConDE & PyData Berlin 2022 Talk "5 Steps to Speed Up Your Data-Analysis on a Single Core"

5 Steps to Speed Up Your Data-Analysis on a Single Core Material for my talk at the PyConDE & PyData Berlin 2022 Description Your data analysis pipeli

Jonathan Striebel 9 Dec 12, 2022
Detectron2-FC a fast construction platform of neural network algorithm based on detectron2

What is Detectron2-FC Detectron2-FC a fast construction platform of neural network algorithm based on detectron2. We have been working hard in two dir

董晋宗 9 Jun 06, 2022
Machine Learning with JAX Tutorials

The purpose of this repo is to make it easy to get started with JAX. It contains my "Machine Learning with JAX" series of tutorials (YouTube videos and Jupyter Notebooks) as well as the content I fou

Aleksa Gordić 372 Dec 28, 2022
One-Shot Neural Ensemble Architecture Search by Diversity-Guided Search Space Shrinking

One-Shot Neural Ensemble Architecture Search by Diversity-Guided Search Space Shrinking This is an official implementation for NEAS presented in CVPR

Multimedia Research 19 Sep 08, 2022
Lightweight mmm - Lightweight (Bayesian) Media Mix Model

Lightweight (Bayesian) Media Mix Model This is not an official Google product. L

Google 342 Jan 03, 2023
PINN Burgers - 1D Burgers equation simulated by PINN

PINN(s): Physics-Informed Neural Network(s) for Burgers equation This is an impl

ShotaDEGUCHI 1 Feb 12, 2022
A small library for doing fluid simulation with neural networks.

Neural Fluid Fields This is a small library for doing fluid simulation with neural fields. Check out our review paper, Neural Fields in Visual Computi

Towaki 23 Jun 23, 2022