https://github.com/mizzao/libmao

math, statistics, randomization, and other useful common grad-school utilities
https://github.com/mizzao/libmao

Last synced: 3 months ago
JSON representation

math, statistics, randomization, and other useful common grad-school utilities

• Host: GitHub
• URL: https://github.com/mizzao/libmao
• Owner: mizzao
• License: mit
• Created: 2012-12-29T20:39:14.000Z (about 12 years ago)
• Default Branch: master
• Last Pushed: 2016-06-09T21:25:49.000Z (over 8 years ago)
• Last Synced: 2023-08-09T01:12:13.725Z (over 1 year ago)
• Language: Java
• Size: 744 KB
• Stars: 13
• Watchers: 5
• Forks: 4
• Open Issues: 3
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
libmao

======

numerical computation, modeling, and machine learning library with

math, statistics, randomization, and other useful utilities

## The highlights

- [Voting rule computation]

  (src/main/java/net/andrewmao/socialchoice/rules): Computation and

  error metrics for many common voting rules as well as those derived

  from more complicated noise models (see below). These were used in

  the AAAI 2013 paper *Better Human Computation Through Principled

  Voting*.

- [Ranking model estimation]

  (src/main/java/net/andrewmao/models/discretechoice): Efficient

  estimation techniques for models on rankings of items, known

  classically as *discrete choice models* and some more recently as

  *random utility models*. The implementation covers many common

  models such as Mallows, Bradley-Terry, Plackett-Luce, and

  Thurstone-Mosteller. In addition, there is a fast parallelized

  implementation of the Monte Carlo EM (MCEM) algorithm from the NIPS

  2012 paper *Random Utility Theory for Social Choice*.

- [HMM inference for repeated games]

  (src/main/java/net/andrewmao/models/games): Hidden Markov Model

  based inference for players' strategies in repeated games. We

  developed this as a technique to analyze experimental data.

## Other useful pieces

- [Random (weighted) subset and stream

  selection](src/main/java/net/andrewmao/math/RandomSelection.java): A

  class for shuffling and selecting random items from a Collection

  efficiently.

- [Truncated Normal and Multivariate Normal

  Distributions](src/main/java/net/andrewmao/probability): Apart from

  a few existing subroutines for normal distributions, two things you

  won't find elsewhere for Java are implementations of truncated

  Normal distributions and a thread-safe library for computing the CDF

  and expected first and second moments of a multivariate normal

  distribution. [Alan

  Genz](http://www.math.wsu.edu/faculty/genz/homepage) deserves credit

  for the actual subroutine, but this code makes it callable from Java

  and thread-safe.

## Interesting tidbits

- [LibraryReplicator]

  (src/main/java/net/andrewmao/misc/LibraryReplicator.java) was a

  class designed to make a non-reentrant native library reentrant. It

  does this by making *n* copies of the library, where *n* is the

  desired level of concurrency, and creating a proxied interface that

  shares the copies of the native library using a thread-safe

  queue. See http://stackoverflow.com/q/14553996/586086 for more

  information on why this was needed.