What I didn't do in GSoC 2018

[ad71]

What I did this summer:
Here is a gist of what I did. It's just an enumeration of the modules/algorithms that I worked on and doesn't go into much detail.

What remains to be done:
Everything apart from the following three algorithms, has been implemented with tests and documentation.

• The HybridWumpusAgent class needs fixing. Some helper functions don't work as expected. The algorithm has no tests and no documentation. Some related helper functions take too long to run and require refactors or re-implementation.
• The cross_validation section needs closure. Refer to this issue, this issue and all linked issues for more details. Tests and documentation is incomplete.
• The DecisionListLearner is not yet complete. Refer to this issue. Tests and documentation is incomplete as well.

What else can be done:

• The project could benefit from GUI apps for some algorithms or standalone visualizations which users can interact with. Here are some examples built using tkinter. I made a couple of these, but quite a few algorithms need a corresponding applet.
• Some algorithms need better and more intuitive explanations in the notebooks. Several notebook sections are pretty drab, and pictures, visualizations, graphs or interactive widgets could help.
• Various algorithms from the learning.py module cannot handle multi-dimensional data (like images) well. If this can be improved, the module can be used as a neural network and pre-processing library for simple tasks and will increase the utility of the project as a whole.
• NLP enthusiasts can try this.
• Also, please refer to the what I didn't do sections of @MrDupin's writeup and @Chipe1's writeup from last year. Some of it still requires attention.

While there are no new algorithms to implement, three of them require refactors to get them working as outlined above. New contributors are welcome to try their hands at improving the project in the ways stated above or can come up with their own ideas as well.

Open to questions and criticism.

[pyaf]

I would like to fix the issues with HybridWumpusAgent, can I take this up?

[ad71]

@pyaf Sure, nobody else is currently working on this issue, please go ahead.

[AbhayGaur]

wow

[KumarArindam]

Wow. Great job.

[ankursikarwar]

I would like to work on an interactive GUI app for the visualization of the backpropagation algorithm from chapter 18.

[ad71]

@ankursikarwar sure, go ahead.

[zeph1yr]

Thanks for your efforts, Is there anything on this list incomplete which I can try to contribute to?

[ad71]

@zeph1yr HybridWumpusAgent isn't fully functional yet, cross-validation is on a bit of a hiatus and decision list-learner has been implemented by @pyaf .
The 'What else...' section is still open for experimentation. If you have a new idea, create an issue if you want to discuss with the collaborators or drop a PR if you're confident.