Summer(?) plans

[dktcoding]

I've been drafting a small plan for my involvement over the summer (summer in Argentina...) in libai (needless to say I'll keep collaborating afterwards, just not so much).

• Make JavaDoc compile without warnings (I'll do this today)
• Add extra activation functions
• Write some integration tests for MLP, SVM and DFS (we'll need a special thread for this, but I think that we can do something relatively good using the current demos)
• Start drafting a Wiki (there's some interest in libai and this will help start things up, again based on the demos)
• Draft and implement a standard way to do filtering and sorting in DataSet
• Generalize a bit TextFileDataSet (tsv, quoted text, etc)
• Export matrices in .bin format

If there's time, add input validation for all the public high level methods (I've been doing some profiling and it really degradates the Matrix class, so there, I'll just make some more verbose asserts)

This are some things that I need, but if there's something else, or something superfluous let me know (my "original" plan was almost 200 lines... but I need to be realistic on my "free-time" expectations)

[kronenthaler]

Sounds good to me man, myself i'm focusing on to finish another project that needed a lot of love, once i'm done with it, i'll spend some time on this one again. I created a project board, so feel free to create issues there, so i can pick up some of them when ready.

I can see your main interest is in the neural networks, if i were you, i would let out any work not related to the NN, like the data sets. I also have some personal interest into improve that implementation, and provide a better integration with the NN package, so it might take some refactors in the near future.

[dktcoding]

You are right, my main interest is NN, but I have been playing around with some of the "other" libai algorithm implementations and well... it all kinda... snowballed.

I'll start working on saving Matrix objects as .bin and .mat files for now to allow integration with GNU Octave/Matlab, then a small break, and then we'll see were we go from there.