sourcecred / research

Repository for research-related items on SourceCred's agenda
9 stars 7 forks source link

unit testing for page_ranker.py #18

Open mzargham opened 5 years ago

mzargham commented 5 years ago

The page_ranker.py file contains methods for imposing heuristics on graphs to create in and out weights in accordance with a large class of pagerank variations including but not limited personal page rank via seed selection and the lazy random walk. This tool is intended to provide a highly generalized method supporting A/B testing and parameter sweeps to hone in on a preferred pagerank algorithm in a scientific manner.

https://github.com/sourcecred/research/blob/mz-exploratory/exploratory/mZargham/pre-infra/page_ranker.py

@mzargham requesting help to set up proper unit tests prior to submitting pull request to master as 'infra'

@BrianLitwin ~ I set up a gitcoin bounty for this. Two ways you can help: 1) add comments with details on how to go about fulfilling this task 2) go ahead and do it yourself and claim the bounty.

gitcoinbot commented 5 years ago

Issue Status: 1. Open 2. Started 3. Submitted 4. Done


This issue now has a funding of 0.5 ETH (86.33 USD @ $172.66/ETH) attached to it.

gitcoinbot commented 5 years ago

Issue Status: 1. Open 2. Cancelled


Work has been started.

These users each claimed they can complete the work by 1 week, 2 days ago. Please review their action plans below:

1) satyaki-mallick has started work.

I am interested to work in this.

Learn more on the Gitcoin Issue Details page.

satyaki-mallick commented 5 years ago

Hi @mzargham , have you gone to any extent on this, the test cases .. ? Anything that you have started working on.

mzargham commented 5 years ago

@satyaki-mallick Ideally we'd match the test cases from the javascript implementation. https://github.com/sourcecred/odyssey-hackathon/blob/master/src/core/pagerankGraph.test.js https://github.com/sourcecred/odyssey-hackathon/blob/master/src/core/attribution/markovChain.test.js To the extent that they apply.

I am a researcher much more so than a dev, but this python method is intended to be a much more verbose version of the pagerank algorithm in js so we can do scientific experiments over the choices alpha, seed, edge weights, etc.

I haven't worked on this file recently; currently trying to get the https://github.com/sourcecred/research/tree/graph-generators-for-testing cleaned up so I can use it alongside the pageranker code for running experiments.

satyaki-mallick commented 5 years ago

Okay. Cool.

gitcoinbot commented 5 years ago

Issue Status: 1. Open 2. Cancelled


The funding of 0.5 ETH (95.0 USD @ $190.01/ETH) attached to this issue has been cancelled by the bounty submitter

mzargham commented 5 years ago

I removed the bounty. Its own me to make sure the Issue is better defined before opening it up to the public. Will continue to use bounties though.