User Alex R.

92 votes

What should I do when my neural network doesn't learn?

answered Jun 19, 2018 at 18:45

43 votes

Accepted

If I generate a random symmetric matrix, what's the chance it is positive definite?

answered Jan 8, 2018 at 19:12

38 votes

Does the normal distribution converge to a uniform distribution when the standard deviation grows to infinity?

answered Jul 10, 2018 at 17:37

33 votes

Does the reciprocal of a probability represent anything?

answered Oct 28, 2016 at 18:59

29 votes

What can one conclude about the data when arithmetic mean is very close to geometric mean?

answered Jun 28, 2016 at 19:59

23 votes

Accepted

Why did statisticians define random matrices?

answered Nov 3, 2016 at 20:15

19 votes

Why is it so important to have principled and mathematical theories for Machine Learning?

answered Dec 12, 2017 at 19:07

16 votes

Accepted

When the Central Limit Theorem and the Law of Large Numbers disagree

answered Jun 25, 2018 at 8:03

14 votes

Iconic (toy) models of neural networks

answered May 15, 2017 at 16:03

13 votes

Accepted

If we primarily use LSTMs over RNNs to solve the vanishing gradient problem, why can't we just use ReLUs/leaky ReLUs with RNNs instead?

answered Apr 3, 2017 at 17:36

12 votes

Why are my steps getting smaller when using fixed step size in gradient descent?

answered Jan 4, 2018 at 20:15

11 votes

Accepted

How can I quickly detect cheating variables in large data?

answered May 5, 2017 at 17:13

11 votes

When is t-SNE misleading?

answered Sep 23, 2016 at 18:45

10 votes

the turning point test

answered Aug 29, 2016 at 18:12

10 votes

Why are symmetric positive definite (SPD) matrices so important?

answered Jul 15, 2016 at 23:31

10 votes

Accepted

What does it mean for a moment generating function to exist in a neighborhood of 0?

answered May 10, 2016 at 1:29

9 votes

Accepted

How to get the confidence interval of a Bernoulli trial if $\hat{p} = 0$?

answered Nov 2, 2016 at 18:57

9 votes

Accepted

How can we prove that a normalized kernel is also a kernel?

answered Nov 21, 2017 at 0:20

9 votes

Accepted

Residual network: why is each block learning residual error with respect to identity mapping?

answered Oct 26, 2017 at 16:49

8 votes

Why do neural networks need feature selection / engineering?

answered May 31, 2018 at 17:43

8 votes

Central Limit Theorem for Markov Chains

answered Nov 3, 2016 at 4:43

8 votes

Value of $\mathbb{E}(\frac{1}{aX+b})$?

answered Nov 30, 2015 at 20:01

8 votes

How do you calculate expected value in this game of Heads or Tails?

answered Dec 2, 2015 at 19:57

8 votes

Accepted

Using the central limit theorem

answered Mar 13, 2016 at 4:22

8 votes

comparison of SGD and ALS in collaborative filtering

answered Mar 23, 2016 at 20:31

8 votes

How does cosine similarity change after a linear transformation?

answered Apr 4, 2016 at 19:50

7 votes

Use of Metropolis & Rejection & Inverse Transform sampling methods

answered Aug 30, 2016 at 21:10

7 votes

Why the default matrix norm is spectral norm and not Frobenius norm?

answered Aug 11, 2016 at 18:25

7 votes

What are the advantages of an exponential random generator using the method of Ahrens and Dieter (1972) rather than by inverse transform?

answered Oct 3, 2016 at 19:02

7 votes

What is an Artificial Neural Network?

answered Aug 16, 2018 at 23:37

Stack Exchange Network

320 Answers

What should I do when my neural network doesn't learn?

If I generate a random symmetric matrix, what's the chance it is positive definite?

Does the normal distribution converge to a uniform distribution when the standard deviation grows to infinity?

Does the reciprocal of a probability represent anything?

What can one conclude about the data when arithmetic mean is very close to geometric mean?

Why did statisticians define random matrices?

Why is it so important to have principled and mathematical theories for Machine Learning?

When the Central Limit Theorem and the Law of Large Numbers disagree

Iconic (toy) models of neural networks

If we primarily use LSTMs over RNNs to solve the vanishing gradient problem, why can't we just use ReLUs/leaky ReLUs with RNNs instead?

Why are my steps getting smaller when using fixed step size in gradient descent?

How can I quickly detect cheating variables in large data?

When is t-SNE misleading?

the turning point test

Why are symmetric positive definite (SPD) matrices so important?

What does it mean for a moment generating function to exist in a neighborhood of 0?

How to get the confidence interval of a Bernoulli trial if $\hat{p} = 0$?

How can we prove that a normalized kernel is also a kernel?

Residual network: why is each block learning residual error with respect to identity mapping?

Why do neural networks need feature selection / engineering?

Central Limit Theorem for Markov Chains

Value of $\mathbb{E}(\frac{1}{aX+b})$?

How do you calculate expected value in this game of Heads or Tails?

Using the central limit theorem

comparison of SGD and ALS in collaborative filtering

How does cosine similarity change after a linear transformation?

Use of Metropolis & Rejection & Inverse Transform sampling methods

Why the default matrix norm is spectral norm and not Frobenius norm?

What are the advantages of an exponential random generator using the method of Ahrens and Dieter (1972) rather than by inverse transform?

What is an Artificial Neural Network?