User stoic-santiago

7 votes

2 answers

4k views

Aren't all discrete convolutions (not just 2D) linear transforms?

asked Mar 30, 2020 at 12:47

6 votes

1 answer

1k views

Why do we update all layers simultaneously while training a neural network?

asked Apr 16, 2020 at 6:37

6 votes

1 answer

157 views

What does the number of required expert demonstrations in Imitation Learning depend on?

asked Aug 13, 2020 at 11:01

6 votes

1 answer

2k views

What are the pros and cons of sparse and dense rewards in reinforcement learning?

asked Aug 13, 2020 at 7:05

5 votes

1 answer

833 views

What's the optimal policy in the rock-paper-scissors game?

asked Aug 27, 2020 at 10:22

5 votes

1 answer

431 views

Why does TD Learning require Markovian domains?

asked Aug 7, 2020 at 5:19

4 votes

1 answer

1k views

When do SARSA and Q-Learning converge to optimal Q values?

asked Aug 9, 2020 at 15:35

4 votes

1 answer

404 views

Why are state-values alone not sufficient in determining a policy (without a model)?

asked Aug 7, 2020 at 3:57

4 votes

1 answer

187 views

Why does REINFORCE work at all?

asked Aug 15, 2020 at 12:30

4 votes

2 answers

667 views

Why is regret so defined in MABs?

asked Sep 26, 2020 at 15:45

4 votes

2 answers

710 views

What is the effect of using pooling layers in CNNs?

asked May 30, 2020 at 10:16

4 votes

0 answers

449 views

How does SGD escape local minima?

asked May 31, 2020 at 9:56

4 votes

1 answer

144 views

Why does L1 regularization yield sparse features?

asked Jul 2, 2020 at 8:44

3 votes

2 answers

4k views

What's the difference between estimation and approximation error?

asked Oct 2, 2020 at 15:48

3 votes

0 answers

32 views

What is meant by subspace clustering in MFA?

asked Oct 21, 2020 at 14:16

3 votes

2 answers

224 views

Why does (not) the distribution of states depend on the policy parameters that induce it?

asked Aug 27, 2020 at 10:36

2 votes

1 answer

238 views

What is the surrogate loss function in imitation learning, and how is it different from the true cost?

asked Aug 13, 2020 at 9:15

2 votes

1 answer

66 views

Is there an equivalent model to the Hidden Markov Model for continuous hidden variables?

asked Oct 21, 2020 at 16:19

2 votes

0 answers

678 views

What's the nearest neighbor algorithm used for upsampling?

asked Aug 20, 2020 at 9:44

2 votes

0 answers

52 views

What are some other real-life examples of simple policies but complex value functions?

asked Aug 27, 2020 at 10:16

2 votes

2 answers

2k views

Why are decision trees and random forests scale invariant?

asked Jul 2, 2020 at 7:55

2 votes

0 answers

72 views

Why is 'scatter' used instead of variance in LDA?

asked Jul 3, 2020 at 5:28

2 votes

1 answer

66 views

How exactly does nested cross-validation work?

asked Jul 5, 2020 at 5:39

2 votes

1 answer

180 views

Why can't pure KG embedding methods discover multi-hop relations paths?

asked May 7, 2020 at 7:25

2 votes

3 answers

556 views

Why does it make sense to study MDPs with finite state and action spaces?

asked Apr 28, 2020 at 12:58

2 votes

0 answers

91 views

How are the classical MDP and the object-oriented MDP views different?

asked Apr 28, 2020 at 14:42

2 votes

1 answer

995 views

What is the KWIK framework?

asked Apr 29, 2020 at 13:57

2 votes

2 answers

138 views

How can Siamese Networks be viewed as RNNs?

asked Apr 30, 2020 at 3:53

2 votes

0 answers

17 views

How do multiple coordinate systems help in capturing invariant features?

asked Oct 20, 2020 at 15:55

1 vote

1 answer

50 views

Why is this variable in equation 2 of the SQAIR paper a random vector of $n$ ones followed by a zero?

asked Apr 30, 2020 at 8:37

Stack Exchange Network

40 Questions

Aren't all discrete convolutions (not just 2D) linear transforms?

Why do we update all layers simultaneously while training a neural network?

What does the number of required expert demonstrations in Imitation Learning depend on?

What are the pros and cons of sparse and dense rewards in reinforcement learning?

What's the optimal policy in the rock-paper-scissors game?

Why does TD Learning require Markovian domains?

When do SARSA and Q-Learning converge to optimal Q values?

Why are state-values alone not sufficient in determining a policy (without a model)?

Why does REINFORCE work at all?

Why is regret so defined in MABs?

What is the effect of using pooling layers in CNNs?

How does SGD escape local minima?

Why does L1 regularization yield sparse features?

What's the difference between estimation and approximation error?

What is meant by subspace clustering in MFA?

Why does (not) the distribution of states depend on the policy parameters that induce it?

What is the surrogate loss function in imitation learning, and how is it different from the true cost?

Is there an equivalent model to the Hidden Markov Model for continuous hidden variables?

What's the nearest neighbor algorithm used for upsampling?

What are some other real-life examples of simple policies but complex value functions?

Why are decision trees and random forests scale invariant?

Why is 'scatter' used instead of variance in LDA?

How exactly does nested cross-validation work?

Why can't pure KG embedding methods discover multi-hop relations paths?

Why does it make sense to study MDPs with finite state and action spaces?

How are the classical MDP and the object-oriented MDP views different?

What is the KWIK framework?

How can Siamese Networks be viewed as RNNs?

How do multiple coordinate systems help in capturing invariant features?

Why is this variable in equation 2 of the SQAIR paper a random vector of $n$ ones followed by a zero?