An open API service indexing awesome lists of open source software.

https://github.com/arranger1044/probabilistic-circuits

A curated collection of papers on probabilistic circuits, computational graphs encoding tractable probability distributions.
https://github.com/arranger1044/probabilistic-circuits

ai ml pc probabilistic-models tractable-inference

Last synced: 5 months ago
JSON representation

A curated collection of papers on probabilistic circuits, computational graphs encoding tractable probability distributions.

Awesome Lists containing this project

README

          

# Probabilistic Circuits

This repo contains the source code for the website [https://arranger1044.github.io/probabilistic-circuits/](https://arranger1044.github.io/probabilistic-circuits/) which is a curated and reasoned list of papers on probabilistic circuits (PCs), computational graphs encoding tractable probability distributions.

## License

[![CC0](http://i.creativecommons.org/p/zero/1.0/88x31.png)](http://creativecommons.org/publicdomain/zero/1.0/)

All the material in this repo is released to the Public Domain. Feel free to clone, fork or complete and/or correct any of these lists.

## How to contribute

To add, change or remove a paper on the website, please open a [pull request](https://github.com/arranger1044/probabilistic-circuits/pulls)!

This site harness Jekyll templates in github pages and their file-based model view. Each paper in the website is associated a markdown file under the `_papers` folder. Modifications to the key-value pairs in this single file would be reflected to the whole website.

Mandatory keys in a paper description are:
- `layout` to be left to `paper`
- `ref` a string acting as a unique identifier
- `title` the complete paper title
- `date` intended as a publication date (only the year matters)
- `tags` a space-separated sequence of tags to classify the paper (see below)
- `authors` a string with authors names, separated by comma
- `venue` the publication venue (conference, journal name)

Optional keys are:
- `pdf` a link to a publicly readable version of the paper
- `code` link to the code released with the paper
- `abstract` the paper abstract, as a single string
- `bibtex` a string for the bibtex entry

The script `dblp_to_md.py` is a quick and dirt way to generate a skeleton of a markdown file entry from the condensed bibtex as available from [DBLP](https://dblp.org/)

### Available tags

Papers on PCs can be catalogued according to the following tags.

Models:
- `acs`: Arithmetic circuits
- `cnets`: Cutset networks
- `spns`: Sum-Product networks
- `aogs`: And/Or graphs
- `pdgs`: Probabilistic decision graphs
- `psdds`: Probabilistic sentential decision diagrams
- `pcs`: Other probabilistic circuits

Algorithms:
- `str-le`: Structure learning
- `par-le`: Parameter learning
- `comp`: Compilation

Inference:
- `mar`: Marginal inference
- `map`: MAP inference
- `mmap`: Marginal MAP inference
- `div`: Divergences, IPMs
- `exp`: Expectations
- `mom`: Moments
- `sam`: Sampling
- `app`: Approximate inference
- `imp`: Imprecise probabilities

Applications:
- `cv`: Computer vision
- `nlp`: Natural language processing
- `seg`: Semantic segmentation
- `act`: Activity recognition
- `spe`: Speech recognition and reconstruction
- `rob`: Robotics
- `bio`: Computational biology
- `the`: Theory
- `ppl`: Probabilistic Programming
- `rep`: Representation Learning
- `hw`: Hardware
- `sw`: Software
- `xai`: Explanations
- `misc`: Other applications

## Thanks

Special thanks to [Giuseppe Lobraico](https://github.com/your) who taught me how to deal with the ruby stack behind Jekyll.