User mshlis

1 vote

When could a linear discriminant give excellent or possibly even the optimal classification accuracy?

answered Aug 12, 2019 at 11:31

1 vote

Accepted

When is the loss calculated, and when does the back-propagation take place?

answered Aug 7, 2019 at 12:23

1 vote

Accepted

What is "Word Sense Disambiguation"?

answered Aug 6, 2019 at 20:14

3 votes

Accepted

What is the difference between asymmetric and depthwise separable convolution?

answered Aug 2, 2019 at 16:56

1 vote

Accepted

Relation between size of parameters and complexity of model with overfitting

answered Aug 2, 2019 at 14:31

3 votes

Is randomness anti-logical?

answered Aug 2, 2019 at 4:51

0 votes

What kind of output should be used for predicting angles in DNNs?

answered Jul 31, 2019 at 19:13

6 votes

Accepted

Is it ok to struggle with mathematics while learning AI as a beginner?

answered Jul 31, 2019 at 13:18

1 vote

Is there a continuous conditional variational auto-encoder?

answered Jul 31, 2019 at 12:49

1 vote

Accepted

What weights should I use while back-propagating?

answered Jul 29, 2019 at 13:04

2 votes

Which approach can I use to generate text based on multiple inputs?

answered Jul 28, 2019 at 17:25

2 votes

Why do language models place less importance on punctuation?

answered Jul 26, 2019 at 14:51

1 vote

What are the benefits of using max-pooling in convolutional neural networks?

answered Jul 25, 2019 at 18:00

2 votes

Accepted

Is adding the Frobenius inner products between filter and input part of convolution or a separate step?

answered Jul 25, 2019 at 1:02

2 votes

Accepted

What is the purpose of the noise injection in the generator network of a GAN?

answered Jul 24, 2019 at 12:26

1 vote

What is the role of the 'fuzzifier' w in Fuzzy Clustering?

answered Jul 23, 2019 at 19:35

1 vote

Accepted

Is PReLU superfluous with respect to ReLU?

answered Jul 23, 2019 at 15:15

8 votes

Accepted

Why can't neural networks learn functions outside of the specified domains?

answered Jul 22, 2019 at 23:00

1 vote

Should I model a problem with quantised output as classification or regression?

answered Jul 22, 2019 at 15:10

2 votes

Are feature maps merged or are they passed on as they are?

answered Jul 20, 2019 at 12:43

0 votes

Accepted

If the goal of training of a GAN is to have $P_g=P_{data}$, shouldn't this produce the exact same images?

answered Jul 19, 2019 at 13:56

1 vote

Accepted

Understanding the reconstruction loss in the paper "Anomaly Detection using Deep Learning based Image Completion"

answered Jul 18, 2019 at 15:13

17 votes

How to stay a up-to-date researcher in ML/RL community?

answered Jul 18, 2019 at 12:23

3 votes

Why do we normalize data in a deep neural network?

answered Jul 17, 2019 at 22:18

1 vote

Is a multi-layer Kohonen network possible?

answered Jul 17, 2019 at 18:16

0 votes

Accepted

How are the parameters of the Bernoulli distribution learned?

answered Jul 17, 2019 at 16:43

0 votes

Can non-Markov environments also be deterministic?

answered Jul 16, 2019 at 4:22

1 vote

How to enforce covariance-matrix output as part of the last layer of a Policy Network?

answered Jul 14, 2019 at 16:29

1 vote

What is the most common practice to apply batch normalization?

answered Jul 11, 2019 at 21:14

1 vote

What is the best loss function for convolution neural network and autoencoder?

answered Jul 11, 2019 at 18:04

Stack Exchange Network

90 Answers

When could a linear discriminant give excellent or possibly even the optimal classification accuracy?

When is the loss calculated, and when does the back-propagation take place?

What is "Word Sense Disambiguation"?

What is the difference between asymmetric and depthwise separable convolution?

Relation between size of parameters and complexity of model with overfitting

Is randomness anti-logical?

What kind of output should be used for predicting angles in DNNs?

Is it ok to struggle with mathematics while learning AI as a beginner?

Is there a continuous conditional variational auto-encoder?

What weights should I use while back-propagating?

Which approach can I use to generate text based on multiple inputs?

Why do language models place less importance on punctuation?

What are the benefits of using max-pooling in convolutional neural networks?

Is adding the Frobenius inner products between filter and input part of convolution or a separate step?

What is the purpose of the noise injection in the generator network of a GAN?

What is the role of the 'fuzzifier' w in Fuzzy Clustering?

Is PReLU superfluous with respect to ReLU?

Why can't neural networks learn functions outside of the specified domains?

Should I model a problem with quantised output as classification or regression?

Are feature maps merged or are they passed on as they are?

If the goal of training of a GAN is to have $P_g=P_{data}$, shouldn't this produce the exact same images?

Understanding the reconstruction loss in the paper "Anomaly Detection using Deep Learning based Image Completion"

How to stay a up-to-date researcher in ML/RL community?

Why do we normalize data in a deep neural network?

Is a multi-layer Kohonen network possible?

How are the parameters of the Bernoulli distribution learned?

Can non-Markov environments also be deterministic?

How to enforce covariance-matrix output as part of the last layer of a Policy Network?

What is the most common practice to apply batch normalization?

What is the best loss function for convolution neural network and autoencoder?