Life In 19x19

I'm guessing that adding new layers would initially cause performance to drop to basically zero, but it would probably train back to something similar to its old performance somewhat quickly. This is similar to pre-training. It's often helpful to pre-train a model on some simple task (e.g ...

Some go knowledge was involved, but indirectly. A winner is determined (by a simple, non-ML portion of the program) at the terminal state of each game. For training the network, the position on the board is given, along with a history of recent moves, and an estimated winning probability given ...

A really common technique in ML is to reduce the "learning rate" as a model starts to converge. And it produces bumps in the model performance exactly like that. So it probably didn't learn specific knowledge at that point, or anything more important than other parts of the learning process, it was ...

My top-down solution.
import functools

@functools.lru_cache(maxsize=None)
def min_coins(target_sum, coins):
if target_sum == 0:
return 0
if target_sum < 0&#58;
return float&#40;'inf'&#41;
return 1 + min&#40;min_coins&#40;target_sum-c, coins&#41; for c in coins&#41 ...

It's best not to think of artificial neural networks as analogous to a brain. They were originally biologically inspired , but their structure and mechanisms are quite different. The inputs and outputs of biological neurons are voltage spikes; artificial neuron activations are continuous, real ...

I can't help but feel that letting the lower right white group become so heavy was a mistake. After B89, I think he should have found a way to sacrifice it (or he should have saved it himself before then.) Granted, it would be a big sacrifice, as it was eventually separating two weak black groups ...

The board is 3.4 sun one-piece masame. The bowls are beautifully figured yaku-sugi cedar. Pictures don't do them justice. The stones are snow grade, size 36. I didn't originally intend to go for such thick stones, but I really liked this pair of bowls and they were quite large and I didn't want them ...

I knew Kaya was supposed to have a distinctive smell, but I didn't realize it would be this intense. My whole place smells sandlewoody.

9. Multiple Accounts
Do not create multiple accounts. Do not create new accounts to circumvent moderator action, conceal your identity, or for other purposes. Additional accounts may be permitted for technical reasons - please contact a moderator or admin beforehand if you believe this applies to ...

EdLee wrote:Related question: of ALL the games out there, other than Go, which ones have a handicap system routinely used (enjoyed) by people ?

At the amateur level, golf and bowling are played regularly with handicaps.

Props to the Zen folks though; it's running on *way* less hardware (2x12 core Xeon server).

The SAME neural net could instead have been taught to play chess or do anything else. The neural net program itself is knowledge neutral.
This is incorrect. The topology and other characteristics of the network is chosen for a particular problem. The dimensions of these networks were built ...

gowan wrote:Based on what I know about AI neural networks I doubt that this program can explain in conceptual terms why the moves it chooses are good.

This program, no. But recurrent and convolutional neural networks are also used for natural language processing so...

If you're interested in how neural networks work, the tutorials for Tensorflow are actually pretty accessible for anyone with basic programming skill (they are in Python).

I don't know the details about what tools Deepmind used here, but Tensorflow is used by machine learning teams at Google.