![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
![[identity profile]](https://www.dreamwidth.org/img/silk/identity/openid.png){kind=small}
This is making the rounds, but it was so cool I just had to jump on the bandwagon: a really impressive analysis of Choose Your Own Adventure books. I would have liked to see a bit more rigorous graph theory analysis, rating them by how balanced their trees are etc, but still just excellent. Somewhere I still have my copy of The Cave of Time, maybe in a box somewhere in Spokane. I should dig that out sometime.
In other news, I'm actually pretty excited by some of the last minute thesis results I got yesterday. Kind of as a lark and kind of to fill space, I decided to look at the distribution of sampling windows sizes selected in the randomized tree generation process. There were some slight trends towards certain sizes and offsets, so I did a complete run where the pool of random samples from which the trees get to choose were biased towards these distributions. And it worked! Not only did the output improve by ~0.5% but frame evaluation times dropped by up to a second. AND the evaluation times dropped in a really interesting way that ties into discussion of frame evaluation times from earlier in the thesis. Not bad for such a simple idea.
In other news, I'm actually pretty excited by some of the last minute thesis results I got yesterday. Kind of as a lark and kind of to fill space, I decided to look at the distribution of sampling windows sizes selected in the randomized tree generation process. There were some slight trends towards certain sizes and offsets, so I did a complete run where the pool of random samples from which the trees get to choose were biased towards these distributions. And it worked! Not only did the output improve by ~0.5% but frame evaluation times dropped by up to a second. AND the evaluation times dropped in a really interesting way that ties into discussion of frame evaluation times from earlier in the thesis. Not bad for such a simple idea.
no subject