Long Time No Read

[jorose]

Hi everyone!

I don't know if any of you remember me, but I'm back! Sorry for my sudden disappearance a year ago, somehow life got in the way and I kind of forgot about my computer chess addiction. I thought I'd write some things here.

What have I been up to?
Aside from non-programming things I worked on my BSc thesis which was a Go engine with ideas from Matthew Lai's Giraffe. AlphaGo got published and kind of outclassed my thesis, but that's life (also I think I found an improvement over Google in an aspect, but I'm getting off topic). Now that that's officially done (only officially, I'm still having fun improving it) I am looking forward to new projects.

What happened to "The Machine", didn't you want to open source it?
Ah, my old engine, I am getting nostalgic just thinking about it. I did and in fact still do. Actually what happened was I thought I would create a windows executable to go along with its release, but that turned out to be a serious pain as for some reason I couldn't get multi-threading to work with it. Then I had a week where I couldn't work on it and then I got sidetracked... Since then I have gotten better at coding in a way people can read and understand and I am not happy with the state of my old code at the moment. My current plan is to rewrite large parts of it in a way in which I am happy, but without changing how it behaves. Ie he should still sacrifice like a madman and be the same strength.

So what are you planning on doing?
I am having fun with neural networks and have been thinking about my old plan of making a human-like engine. These ideas can combine quite well by utilizing large enough networks to en corporate fine strategic information while simultaneously becoming too slow to allow the engine to be too powerful.

I am starting work on an engine based on the Caffe Deep Learning Framework. I intend to write most of the engine from scratch, though I might copy some of my move generation from my old engine, since I find that part of engine programming rather boring and tedious. Eventually, I want to open source this new engine with different trained neural network models and code that makes it rather easy to generate new models, that way, if someone wants to train an engine based on Morphy's games, they can!

I don't intend to explicitly support Windows as I am moving more and more towards uni-boot Linux systems, but if someone would like to make a Windows binary after release, I would gladly accept help and give credit.