search

pgrew / mbb

🤰🏽 make 🤰🏿 build 🤰🏻 brain
0 stars 1 forks source link

Provide guidance on contributing to this package #9

Open buddha314 opened 7 years ago

buddha314 commented 7 years ago

If this package is going to expand beyond SVMs, I'd be interested in contributing pseudocode, maybe some prototypes and snarky remarks.

I am currently developing a lame prototype for "Modified Adsorption", a Graph-Based Semi-Supervised Learning routine provided on page 93 of this presentation

If this is interesting, I may also be able to provide some developer resources. I'd be happy to type up the algorithm in a Jupyter Notebook. Please let me know what you think.

ben-albrecht commented 7 years ago

Hi @buddha314 --

As seen in the other issues, we are currently porting Parallel-SVM to Chapel and comparing performance of distributed implementations. We're open to other ideas after that project.

Feel free to post ideas as issues or submit example codes as pull requests into the examples/<topic> directory.

pgrew commented 7 years ago

@buddha314 Growing beyond SVMs is exactly what we want! I would be thrilled to read your algorithm.

ben-albrecht commented 7 years ago

There may be some overlap in the goals of this package and numsuch. It might be worth looking into ways to combine efforts, once this repository expands beyond SVMs.

buddha314 commented 7 years ago

I have crappy Multi-Layer Perceptron here and and about 80% into Modified Adsorption here.

I have no desire to own the project for very long, but here are some things I think should persist once younger, smarter folk take over.

  1. The name is easy to remember and does a Python-like play on the language name.
  2. A Keras-like and Scipy-like interface. model.add() is pretty cool
  3. Mason compatibility
  4. I want to be able to contribute the things I need.

So my suggestion is: Once I get done with the MAD branch, I am happy to take the passenger seat. You guys wanna drive?