3 ms·
Yeah it does, but from the dynamics side I mentioned. How would you find enough data to train such a model? AlphaGo and the StarCraft thing were trained on rid
by lordnacho 6y ago
Yeah it does, but from the dynamics side I mentioned. How would you find enough data to train such a model?
AlphaGo and the StarCraft thing were trained on ridiculously huge datasets IIUC.
- pixelpoet 6y agoAlphaGo (or rather AlphaZero) is a bit different because it trains from self-play, where you can get a(n unbiased approximation to the) true answer via Monte Carlo Tree Search.