PolytopeSampler is a Matlab implementation of constrained Riemannian Hamiltonian Monte Carlo for sampling from high dimensional disributions on polytopes

Overview

PolytopeSampler

PolytopeSampler is a Matlab implementation of constrained Riemannian Hamiltonian Monte Carlo for sampling from high dimensional disributions on polytopes. It is able to sample efficiently from sets and distributions with more than 100K dimensions.

Quick Tutorial

PolytopeSampler samples from distributions of the form exp(-f(x)), for a convex function f, subject to constraints Aineq * x <= bineq, Aeq * x = beq and lb <= x <= ub.

The function f can be specified by arrays containing its first and second derivative or function handles. Only the first derivative is required. By default, f is empty, which represents a uniform distribution. If the first derivative is a function handle, then the function and its second derivatives must also be provided.

To sample N points from a polytope P, you can call sample(P, N). The function sample will

  1. Find an initial feasible point
  2. Run constrained Hamiltonian Monte Carlo
  3. Test convergence of the sampling algorithm by computing Effective Sample Size (ESS) and terminate when ESS >= N. If the target distribution is uniform, a uniformity test will also be performed.

Extra parameters can be set up using opts. Some useful parameters include maxTime and maxStep. By default, they are set to

                        maxTime: 86400 (max sampling time in seconds)
                        maxStep: 300000 (maximum number of steps)

The output is a struct o, which stores samples generated in o.samples and a summary of the sample in o.summary. o.samples is an array of size dim x #steps.

Example

We demonstrate PolytopeSampler using a simple example, sampling uniformly from a simplex. The polytope is defined by

>> P = struct;
>> d = 10;
>> P.Aeq = ones(1, d);
>> P.beq = 1;
>> P.lb = zeros(d, 1);

The polytope has dimension d = 10 with constraint sum_i x_i = 1 and x >= 0. This is a simplex. To generate 200 samples uniformly from the polytope P, we call the function sample().

>> o = sample(P, 200);
  Time spent |  Time reamin |                  Progress | Samples |  AccProb | StepSize |  MixTime
00d:00:00:01 | 00d:00:00:00 | ######################### | 211/200 | 0.989903 | 0.200000 |     11.2
Done!

We can access the samples generated using

>> o.samples

We can print a summary of the samples:

>> o.summary

ans =

  10×7 table

                     mean        std         25%         50%         75%      n_ess      r_hat 
                   ________    ________    ________    ________    _______    ______    _______

    samples[1]     0.093187    0.091207    0.026222    0.064326    0.13375    221.51    0.99954
    samples[2]     0.092815    0.086905    0.027018    0.066017    0.13221    234.59     1.0301
    samples[3]      0.10034    0.090834    0.030968    0.075631    0.13788    216.56     1.0159
    samples[4]      0.10531    0.092285    0.035363    0.077519     0.1481    235.25     1.0062
    samples[5]      0.10437    0.087634    0.034946    0.080095     0.1533    212.54    0.99841
    samples[6]       0.1029    0.093724    0.028774    0.074354    0.15135     227.6     1.0052
    samples[7]       0.1042    0.083084    0.038431    0.081964    0.15352    231.54     1.0008
    samples[8]     0.088778    0.086902    0.025565    0.062473    0.11837    229.69     1.0469
    samples[9]      0.10627     0.09074    0.036962    0.084294    0.15125    211.64    0.99856
    samples[10]     0.10184    0.084699    0.035981    0.074923    0.14578    230.63     1.0277

n_ess shows the effective sample size of the samples generated. r_hat tests the convergence of the sampling algorithm. A value of r_hat close to 1 indicates that the algorithm has converged properly.

See demo.m for more examples, including examples of sampling from non-uniform distributions.

A shimmer pre-load component for Plotly Dash

dash-loading-shimmer A shimmer pre-load component for Plotly Dash Installation Get it with pip: pip install dash-loading-extras Or maybe you prefer Pi

Lucas Durand 4 Oct 12, 2022
Advanced hot reloading for Python

The missing element of Python - Advanced Hot Reloading Details Reloadium adds hot reloading also called "edit and continue" functionality to any Pytho

Reloadware 1.9k Jan 04, 2023
HiPlot makes understanding high dimensional data easy

HiPlot - High dimensional Interactive Plotting HiPlot is a lightweight interactive visualization tool to help AI researchers discover correlations and

Facebook Research 2.4k Jan 04, 2023
High performance, editable, stylable datagrids in jupyter and jupyterlab

An ipywidgets wrapper of regular-table for Jupyter. Examples Two Billion Rows Notebook Click Events Notebook Edit Events Notebook Styling Notebook Pan

J.P. Morgan Chase 75 Dec 15, 2022
Mattia Ficarelli 2 Mar 29, 2022
Data visualization using matplotlib

Data visualization using matplotlib project instructions Top 5 Most Common Coffee Origins In this visualization I used data from Ankur Chavda on Kaggl

13 Oct 27, 2021
Process dataframe in a easily way.

Popanda Written by Shengxuan Wang at OSU. Used for processing dataframe, especially for machine learning. The name is from "Po" in the movie Kung Fu P

ShawnWang 1 Dec 24, 2021
A programming language built on top of Python to easily allow Swahili speakers to get started with programming without ever knowing English

pyswahili A programming language built over Python to easily allow swahili speakers to get started with programming without ever knowing english pyswa

Jordan Kalebu 72 Dec 15, 2022
Visualizations for machine learning datasets

Introduction The facets project contains two visualizations for understanding and analyzing machine learning datasets: Facets Overview and Facets Dive

PAIR code 7.1k Jan 07, 2023
A tool for automatically generating 3D printable STLs from freely available lidar scan data.

mini-map-maker A tool for automatically generating 3D printable STLs from freely available lidar scan data. Screenshots Tutorial To use this script, g

Mike Abbott 51 Nov 06, 2022
Practical-statistics-for-data-scientists - Code repository for O'Reilly book

Code repository Practical Statistics for Data Scientists: 50+ Essential Concepts Using R and Python by Peter Bruce, Andrew Bruce, and Peter Gedeck Pub

1.7k Jan 04, 2023
Homework 2: Matplotlib and Data Visualization

Homework 2: Matplotlib and Data Visualization Overview These data visualizations were created for my introductory computer science course using Python

Sophia Huang 12 Oct 20, 2022
A small collection of tools made by me, that you can use to visualize atomic orbitals in both 2D and 3D in different aspects.

Orbitals in Python A small collection of tools made by me, that you can use to visualize atomic orbitals in both 2D and 3D in different aspects, and o

Prakrisht Dahiya 1 Nov 25, 2021
Python ts2vg package provides high-performance algorithm implementations to build visibility graphs from time series data.

ts2vg: Time series to visibility graphs The Python ts2vg package provides high-performance algorithm implementations to build visibility graphs from t

Carlos Bergillos 26 Dec 17, 2022
3D plotting and mesh analysis through a streamlined interface for the Visualization Toolkit (VTK)

PyVista Deployment Build Status Metrics Citation License Community 3D plotting and mesh analysis through a streamlined interface for the Visualization

PyVista 1.6k Jan 08, 2023
A simple agent-based model used to teach the basics of OOP in my lectures

Pydemic A simple agent-based model of a pandemic. This is used to teach basic principles of object-oriented programming to master students. It is not

Fabien Maussion 2 Jun 08, 2022
MPL Plotter is a Matplotlib based Python plotting library built with the goal of delivering publication-quality plots concisely.

MPL Plotter is a Matplotlib based Python plotting library built with the goal of delivering publication-quality plots concisely.

Antonio López Rivera 162 Nov 11, 2022
Python library that makes it easy for data scientists to create charts.

Chartify Chartify is a Python library that makes it easy for data scientists to create charts. Why use Chartify? Consistent input data format: Spend l

Spotify 3.2k Jan 04, 2023
Uniform Manifold Approximation and Projection

UMAP Uniform Manifold Approximation and Projection (UMAP) is a dimension reduction technique that can be used for visualisation similarly to t-SNE, bu

Leland McInnes 6k Jan 08, 2023
Library for exploring and validating machine learning data

TensorFlow Data Validation TensorFlow Data Validation (TFDV) is a library for exploring and validating machine learning data. It is designed to be hig

688 Jan 03, 2023