7 ms·

What that tells me is that they haven't yet come up with the right featurization - that is, the function that maps input data into the actual neural network nod

What that tells me is that they haven't yet come up with the right featurization - that is, the function that maps input data into the actual neural network node values. The appropriate featurization would include the time information but reduce its dimensionality by hard-coding some basic assumptions, of the kind that humans presumably make when processing the same data.

yingxie3 9y ago
I think these guys (and most people using deep models) try to avoid hand-crafted features as much as possible.
- eutectic 9y ago
  Alphago used several hand-crafted features as of the Nature paper, so DeepMind at least is not above a little feature engineering.
- kastnerkyle 9y ago
  Gabriel (as well as the others on the team) have definitely looked at these areas - if things were left out/not "featurized" it was likely done via an ablation test, or showed improvement over benchmarks, or maybe just to set a baseline, as he is quoted in the main article. I don't know what techniques they used here, but I am excited to find out! On the specific issue of encoding time-dependent behaviors in models, I think it is related to a broader issue that shows up in many application areas. To me the critical factor is that these models are ruthlessly good at exploiting local dependencies and totally forgetting long-term global dependencies or respecting required structure in control/generation. This basically means it is very difficult to train long-term, time dependent behavior without tricks (early/mid/late game models, extensive handcrafting of the inputs, or using high level "macro actions"). Indeed, FAIR's recent mini-RTS engine ELF directly gives macro actions, in part to look closer at how well global strategies are really handled and remove one factor of complexity [0]. Gabriel's PhD thesis was entirely on Bayesian models for RTS AI, applied to SC:BW [1], so I am sure he is well aware of the "classic/rules based" approaches for this. [0] https://code.facebook.com/posts/132985767285406/introducing-elf-an-extensive-lightweight-and-flexible-platform-for-game-research/ https://code.facebook.com/posts/132985767285406/introducing-... [1] http://emotion.inrialpes.fr/people/synnaeve/phdthesis/phdthesis.html http://emotion.inrialpes.fr/people/synnaeve/phdthesis/phdthe...

yingxie3 9y ago

I think these guys (and most people using deep models) try to avoid hand-crafted features as much as possible.

eutectic 9y ago
Alphago used several hand-crafted features as of the Nature paper, so DeepMind at least is not above a little feature engineering.
kastnerkyle 9y ago
Gabriel (as well as the others on the team) have definitely looked at these areas - if things were left out/not "featurized" it was likely done via an ablation test, or showed improvement over benchmarks, or maybe just to set a baseline, as he is quoted in the main article. I don't know what techniques they used here, but I am excited to find out! On the specific issue of encoding time-dependent behaviors in models, I think it is related to a broader issue that shows up in many application areas. To me the critical factor is that these models are ruthlessly good at exploiting local dependencies and totally forgetting long-term global dependencies or respecting required structure in control/generation. This basically means it is very difficult to train long-term, time dependent behavior without tricks (early/mid/late game models, extensive handcrafting of the inputs, or using high level "macro actions"). Indeed, FAIR's recent mini-RTS engine ELF directly gives macro actions, in part to look closer at how well global strategies are really handled and remove one factor of complexity [0]. Gabriel's PhD thesis was entirely on Bayesian models for RTS AI, applied to SC:BW [1], so I am sure he is well aware of the "classic/rules based" approaches for this. [0] https://code.facebook.com/posts/132985767285406/introducing-elf-an-extensive-lightweight-and-flexible-platform-for-game-research/ https://code.facebook.com/posts/132985767285406/introducing-... [1] http://emotion.inrialpes.fr/people/synnaeve/phdthesis/phdthesis.html http://emotion.inrialpes.fr/people/synnaeve/phdthesis/phdthe...