Most Popular

1500 questions
11
votes
2 answers

How do we prove the n-step return error reduction property?

In section 7.1 (about the n-step bootstrapping) of the book Reinforcement Learning: An Introduction (2nd edition), by Andrew Barto and Richard S. Sutton, the authors write about what they call the "n-step return error reduction property": But they…
123learn
  • 111
  • 3
11
votes
1 answer

What are ontologies in AI?

What exactly are ontologies in AI? How should I write them and why are they important?
oren revenge
  • 253
  • 2
  • 8
11
votes
5 answers

Are methods of exhaustive search considered to be AI?

Some programs do exhaustive searches for a solution while others do heuristic searches for a similar answer. For example, in chess, the search for the best next move tends to be more exhaustive in nature whereas, in Go, the search for the best next…
WilliamKF
  • 2,513
  • 1
  • 25
  • 31
11
votes
1 answer

Can layers of deep neural networks be seen as Hopfield networks?

Hopfield networks are able to store a vector and retrieve it starting from a noisy version of it. They do so setting weights in order to minimize the energy function when all neurons are set equal to the vector values, and retrieve the vector using…
Mario Alemi
  • 211
  • 1
  • 3
11
votes
3 answers

How can AI researchers avoid "overfitting" to commonly-used benchmarks as a community?

In fields such as Machine Learning, we typically (somewhat informally) say that we are overfitting if improve our performance on a training set at the cost of reduced performance on a test set / the true population from which data is sampled. More…
Dennis Soemers
  • 10,264
  • 2
  • 26
  • 69
11
votes
2 answers

What tools are used to deal with adversarial examples problem?

The problem of adversarial examples is known to be critical for neural networks. For example, an image classifier can be manipulated by additively superimposing a different low amplitude image to each of many training examples that looks like noise…
Ilya Palachev
  • 299
  • 2
  • 11
11
votes
2 answers

Why do we prefer ReLU over linear activation functions?

The ReLU activation function is defined as follows $$y = \operatorname{max}(0,x)$$ And the linear activation function is defined as follows $$y = x$$ The ReLU nonlinearity just clips the values less than 0 to 0 and passes everything else. Then why…
imflash217
  • 499
  • 5
  • 14
11
votes
5 answers

What kind of simulated environment is complex enough to develop a general AI?

Imagine trying to create a simulated virtual environment that is complicated enough to create a "general AI" (which I define as a self aware AI) but is as simple as possible. What would this minimal environment be like? i.e. An environment that was…
zooby
  • 2,206
  • 1
  • 13
  • 22
11
votes
1 answer

What kind of problems require more than 2 hidden layers?

I've read that the most of the problems can be solved with 1-2 hidden layers. How do you know you need more than 2? For what kind of problems you would need them (give me an example)?
kenorb
  • 10,483
  • 3
  • 44
  • 94
11
votes
3 answers

What are the mathematical prerequisites to be able to study artificial general intelligence?

What are the mathematical prerequisites to be able to study artificial general intelligence (AGI) or strong AI?
Mark ellon
  • 499
  • 1
  • 5
  • 6
11
votes
2 answers

What are some alternatives to the book "Artificial Intelligence: A Modern Approach"?

There are two textbooks that I most love and am most afraid of in the world: Introduction to Algorithms by Cormen et al. and Artificial Intelligence: A Modern Approach by Norvig et al. I have started the "AI: A Modern Approach" more than once, but…
ml_nrd
  • 111
  • 1
  • 5
11
votes
4 answers

What are some tactics for recognizing artificially made media?

With the growing ability to cheaply create fake pictures, fake soundbites, and fake video there becomes an increasing problem with recognizing what is real and what isn't. Even now we see a number of examples of applications that create fake media…
Andrew Butler
  • 587
  • 3
  • 10
11
votes
3 answers

Why were Chess experts surprised by the AlphaZero's victory against Stockfish?

It was recently brought to my attention that Chess experts took the outcome of this now famous match as something of an upset. See: Chess’s New Best Player Is A Fearless, Swashbuckling Algorithm As as a non-expert on Chess and Chess AI, my…
DukeZhou
  • 6,227
  • 5
  • 25
  • 53
11
votes
1 answer

Do off-policy policy gradient methods exist?

Do off-policy policy gradient methods exist? I know that policy gradient methods themselves using the policy function for sampling rollouts. But can't we easily have a model for sampling from the environment? If so, I've never seen this done before.
echo
  • 693
  • 1
  • 6
  • 12
11
votes
3 answers

What size of neural networks can be trained on current consumer grade GPUs? (1060,1070,1080)

Is it possible to give a rule of thumb estimate about the size of neural networks that are trainable on common consumer-grade GPUs? For example, the Emergence of Locomotion (Reinforcement) paper trains a network using tanh activation of the neurons.…
pascalwhoop
  • 305
  • 1
  • 8