User Benoit Sanchez

47 votes

Accepted

What is a manifold?

terminology manifold-learning

answered Jul 8, 2017 at 11:05

37 votes

Accepted

Is ridge regression useless in high dimensions ($n \ll p$)? How can OLS fail to overfit?

answered Feb 15, 2018 at 9:24

20 votes

Is unbiased maximum likelihood estimator always the best unbiased estimator?

mathematical-statistics maximum-likelihood unbiased-estimator

answered Jan 10, 2018 at 17:07

18 votes

Accepted

How does batch size affect convergence of SGD and why?

machine-learning neural-networks optimization gradient-descent stochastic-gradient-descent

answered Nov 30, 2017 at 19:35

18 votes

Accepted

Relationships between correlation and causation

correlation causality

answered Jun 2, 2017 at 9:36

15 votes

Reduce Classification Probability Threshold

machine-learning classification binary-data threshold

answered Nov 6, 2017 at 15:22

14 votes

How to correctly compute mutual information (Python Example)?

python count-data entropy mutual-information

answered Oct 3, 2017 at 20:57

14 votes

Accepted

K-nearest-neighbour with continuous and binary variables

r classification k-nearest-neighbour

answered Mar 31, 2017 at 16:04

13 votes

Accepted

Can log likelihood funcion be positive

expectation-maximization mixture-distribution likelihood

answered Dec 21, 2017 at 11:07

11 votes

Accepted

How do Bayesian Statistics handle the absence of priors?

bayesian prior mathematical-statistics philosophical

answered Feb 3, 2018 at 13:23

11 votes

Accepted

Negative probabilities: layman explanations

probability

answered Apr 11, 2018 at 10:14

11 votes

How to understand the drawbacks of K-means

machine-learning clustering data-mining k-means

answered Jun 13, 2017 at 12:34

8 votes

Accepted

Usefulness of convexity of linear regression when there is no closed form solution

regression mathematical-statistics optimization matrix-inverse

answered Jul 31, 2017 at 10:15

8 votes

Accepted

Conditional probability formula proof

conditional-probability

answered Apr 5, 2018 at 11:01

7 votes

Accepted

theoretical basis for logistic regression

regression logistic binary-data mathematical-statistics

answered Dec 28, 2017 at 17:20

7 votes

gradient versus partial derivatives

machine-learning gradient-descent calculus

answered Jun 16, 2017 at 11:09

7 votes

A job interview question on flipping a coin

bias

answered May 31, 2017 at 17:02

6 votes

Why do we assume the exponential family in the GLM context?

generalized-linear-model exponential-family

answered Jun 8, 2017 at 15:32

6 votes

how can I convert a negative log likelihood to likelihood?

probability likelihood hidden-markov-model

answered Jul 18, 2017 at 15:18

6 votes

Accepted

Relation between MSE and Bias-Variance

mse bias-variance-tradeoff

answered Dec 29, 2017 at 13:49

6 votes

Why are contours of a multivariate Gaussian distribution elliptical?

normal-distribution multivariate-normal-distribution

answered Feb 1, 2018 at 21:09

6 votes

Accepted

How could I have discovered the normal distribution?

probability distributions normal-distribution references history

answered Feb 9, 2018 at 10:32

5 votes

Investigating differences between populations

anova random-forest cart group-differences

answered Jul 18, 2017 at 23:07

5 votes

Why would you multiply the regression coefficient by the standard deviation of the independent variable?

interpretation regression-coefficients

answered Jun 1, 2017 at 16:10

5 votes

Difference between logit and probit models

r generalized-linear-model logistic probit link-function

answered Jun 10, 2017 at 12:25

5 votes

P_values and effect size

p-value effect-size statistical-power

answered Jun 12, 2017 at 16:38

5 votes

Why is the likelihood function "not the same as the probability that those parameters are the right ones, given an observed sample"?

bayesian mathematical-statistics maximum-likelihood likelihood

answered Jun 24, 2017 at 10:50

4 votes

Accepted

Power of the chi-squared test

mathematical-statistics chi-squared-test goodness-of-fit statistical-power

answered Jul 4, 2017 at 16:00

4 votes

Likelihood in Linear Regression

regression probability linear-model likelihood linear

answered Oct 2, 2017 at 13:06

4 votes

Accepted

Is there a "binding" between loss functions and models?

least-squares loss-functions cross-entropy

answered Oct 9, 2017 at 16:36

Stack Exchange Network

current community

your communities

more stack exchange communities

140 Answers

What is a manifold?

Is ridge regression useless in high dimensions ($n \ll p$)? How can OLS fail to overfit?

Is unbiased maximum likelihood estimator always the best unbiased estimator?

How does batch size affect convergence of SGD and why?

Relationships between correlation and causation

Reduce Classification Probability Threshold

How to correctly compute mutual information (Python Example)?

K-nearest-neighbour with continuous and binary variables

Can log likelihood funcion be positive

How do Bayesian Statistics handle the absence of priors?

Negative probabilities: layman explanations

How to understand the drawbacks of K-means

Usefulness of convexity of linear regression when there is no closed form solution

Conditional probability formula proof

theoretical basis for logistic regression

gradient versus partial derivatives

A job interview question on flipping a coin

Why do we assume the exponential family in the GLM context?

how can I convert a negative log likelihood to likelihood?

Relation between MSE and Bias-Variance

Why are contours of a multivariate Gaussian distribution elliptical?

How could I have discovered the normal distribution?

Investigating differences between populations

Why would you multiply the regression coefficient by the standard deviation of the independent variable?

Difference between logit and probit models

P_values and effect size

Why is the likelihood function "not the same as the probability that those parameters are the right ones, given an observed sample"?

Power of the chi-squared test

Likelihood in Linear Regression

Is there a "binding" between loss functions and models?