User Sobi

30 votes

Accepted

The paradox of i.i.d. data (at least for me)

answered Dec 13, 2015 at 6:24

23 votes

Accepted

Why is the bias term in SVM estimated separately, instead of an extra dimension in the feature vector?

answered Dec 9, 2015 at 17:46

22 votes

Why is the Expectation Maximization algorithm guaranteed to converge to a local optimum?

answered Nov 24, 2015 at 19:30

13 votes

Accepted

What's the relationship between an SVM and hinge loss?

answered Dec 22, 2015 at 15:59

12 votes

Accepted

Can neural network (e.g., convolutional neural network) have negative weights?

answered Dec 9, 2015 at 22:35

10 votes

Accepted

KL divergence and expectations

answered Dec 17, 2015 at 21:49

5 votes

Accepted

Using clustering for unsupervised classification (visualizing k-means cluster centers)

answered Dec 14, 2015 at 3:42

4 votes

How to generate samples uniformly at random from multiple discrete variables subject to constraints?

answered Dec 1, 2015 at 22:13

4 votes

Accepted

What is the motivation for the entropy term in the proof of EM algorithm?

answered Dec 23, 2015 at 21:52

4 votes

Accepted

How to calculate prior probabilities

answered Jan 6, 2016 at 0:49

4 votes

Accepted

Understanding Monte Carlo sampling

answered Dec 15, 2015 at 16:28

4 votes

Accepted

Practical Bayes Theorem - how to use it?

answered Dec 11, 2015 at 19:47

3 votes

Accepted

Hidden Markov Model and Viterbi algorithm: Understanding the Casino Problem?

answered Dec 6, 2015 at 9:10

3 votes

Accepted

Subset selection / feature selection with a categorical response variable

answered Dec 10, 2015 at 20:56

3 votes

Theoretical justification for bag of words

answered Dec 19, 2015 at 4:27

3 votes

Accepted

Visualize a lot of words

answered Dec 19, 2015 at 20:01

2 votes

Accepted

Weights in LinearSVC change if we magnify features

answered Dec 14, 2015 at 23:16

2 votes

Constructing a problem-specific loss function

answered Dec 2, 2015 at 19:02

2 votes

What is the difference between a neural network and a deep neural network, and why do the deep ones work better?

answered Dec 3, 2015 at 21:39

1 vote

how to fit pdf of known form to data

answered Nov 24, 2015 at 23:29

1 vote

Training a basic Markov Random Field for classifying pixels in an image

answered Nov 25, 2015 at 23:12

1 vote

Accuracy reduced with Adaboost

answered Dec 8, 2015 at 7:55

1 vote

Accepted

Advantage of latent SVM for part-based object detection

answered Dec 9, 2015 at 22:01

1 vote

Accepted

Binary linear SVM in terms of structured SVM

answered Dec 10, 2015 at 22:07

1 vote

Efficient computation of the distribution of a weighted sum of finite discrete random variables

answered Dec 20, 2015 at 21:53

1 vote

Accepted

$L_2$ regularization parameter

answered Dec 16, 2015 at 22:04

0 votes

Accepted

How to calculate accuracy of identification using embedings and corresponded labels?

answered Dec 28, 2017 at 16:42

Stack Exchange Network

27 Answers

The paradox of i.i.d. data (at least for me)

Why is the bias term in SVM estimated separately, instead of an extra dimension in the feature vector?

Why is the Expectation Maximization algorithm guaranteed to converge to a local optimum?

What's the relationship between an SVM and hinge loss?

Can neural network (e.g., convolutional neural network) have negative weights?

KL divergence and expectations

Using clustering for unsupervised classification (visualizing k-means cluster centers)

How to generate samples uniformly at random from multiple discrete variables subject to constraints?

What is the motivation for the entropy term in the proof of EM algorithm?

How to calculate prior probabilities

Understanding Monte Carlo sampling

Practical Bayes Theorem - how to use it?

Hidden Markov Model and Viterbi algorithm: Understanding the Casino Problem?

Subset selection / feature selection with a categorical response variable

Theoretical justification for bag of words

Visualize a lot of words

Weights in LinearSVC change if we magnify features

Constructing a problem-specific loss function

What is the difference between a neural network and a deep neural network, and why do the deep ones work better?

how to fit pdf of known form to data

Training a basic Markov Random Field for classifying pixels in an image

Accuracy reduced with Adaboost

Advantage of latent SVM for part-based object detection

Binary linear SVM in terms of structured SVM

Efficient computation of the distribution of a weighted sum of finite discrete random variables

$L_2$ regularization parameter

How to calculate accuracy of identification using embedings and corresponded labels?