User Djib2011

106 votes

Accepted

How does a simple logistic regression model achieve a 92% classification accuracy on MNIST?

answered Sep 11, 2019 at 23:23

84 votes

Accepted

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

answered Sep 7, 2018 at 11:07

25 votes

Accepted

Does the image format (png, jpg, gif) affect how an image recognition neural net is trained?

answered Jun 29, 2017 at 11:48

20 votes

Accepted

Is it 'fair' to set a seed in a random forest regression to yield the highest accuracy?

answered Apr 19, 2018 at 22:34

20 votes

Accepted

How to understand Generative Adversarial Networks Discriminative distribution?

answered Nov 8, 2019 at 10:02

18 votes

Accepted

Difference between Naive Bayes vs Recurrent Neural Network (LSTM)

answered Jun 15, 2017 at 14:39

16 votes

In a convolutional neural network (CNN), when convolving the image, is the operation used the dot product or the sum of element-wise multiplication?

answered Mar 19, 2018 at 2:24

15 votes

Accepted

Normalization/standardization: Should one do this before oversampling/undersampling the data or after?

answered Aug 21, 2018 at 22:21

15 votes

Accepted

Which type of data normalizing should be used with KNN?

answered Aug 25, 2018 at 11:40

12 votes

Why the name "kernel" in stats and ML?

answered Apr 20, 2018 at 22:55

11 votes

Accepted

Why is the $x$ axis for a histogram labeled "bin"?

answered Apr 12, 2018 at 21:44

10 votes

Do I need 3 RGB channels for a spectrogram CNN?

answered Sep 11, 2019 at 17:03

8 votes

Accepted

Does it make sense to use an Early Stopping Metric like “mae” instaed of “val_loss” for regression problems?

answered Aug 13, 2018 at 12:27

7 votes

Accepted

Machine learning methods for multi-dimensional input and output

answered Aug 9, 2018 at 3:48

7 votes

Accepted

Data augmentation or weighted loss function for imbalanced classes?

answered Aug 24, 2018 at 16:43

6 votes

Accepted

Does class balancing introduce bias?

answered Sep 28, 2018 at 23:48

6 votes

Accepted

How to select the best classifier in classification task?

answered Apr 19, 2018 at 22:24

5 votes

Accepted

Link between dimensionality of feature vectors and estimation

answered Aug 5, 2018 at 0:17

5 votes

When we need fine tuning of CNN?

answered Aug 26, 2018 at 12:12

5 votes

Accepted

What is the loss function used for CNN?

answered Oct 24, 2019 at 7:03

4 votes

Accepted

Could ResNet curve be used for a regression problem, e.g. housing price prediction?

answered Mar 27, 2020 at 23:08

4 votes

Accepted

Why is multicollinearity so bad for machine learning models and what can we do about it?

answered Aug 7, 2018 at 0:26

3 votes

How initializing weights to a certain variance reduce vanishing / exploding gradients?

answered Aug 1, 2018 at 10:17

3 votes

When training a neural network to determine appropriate number of epochs should you look at validation loss or validation objective loss?

answered Aug 30, 2018 at 20:20

3 votes

Does it makes sense to perturb images on-the-run while training CNNs?

answered Aug 20, 2018 at 0:04

3 votes

Accepted

What influences fluctuations in validation accuracy?

answered Aug 23, 2018 at 13:09

3 votes

Stopping criteria for stochastic gradient descent?

answered Oct 26, 2019 at 8:48

3 votes

Normalizations: dividing by maximum

answered Jun 14, 2019 at 23:38

3 votes

Accepted

Why CNNs are less prone to overfitting?

answered Jun 22, 2019 at 11:07

3 votes

Accepted

Purpose of Nested Cross-Validation?

answered Mar 2, 2021 at 9:07

Stack Exchange Network

94 Answers

How does a simple logistic regression model achieve a 92% classification accuracy on MNIST?

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

Does the image format (png, jpg, gif) affect how an image recognition neural net is trained?

Is it 'fair' to set a seed in a random forest regression to yield the highest accuracy?

How to understand Generative Adversarial Networks Discriminative distribution?

Difference between Naive Bayes vs Recurrent Neural Network (LSTM)

In a convolutional neural network (CNN), when convolving the image, is the operation used the dot product or the sum of element-wise multiplication?

Normalization/standardization: Should one do this before oversampling/undersampling the data or after?

Which type of data normalizing should be used with KNN?

Why the name "kernel" in stats and ML?

Why is the $x$ axis for a histogram labeled "bin"?

Do I need 3 RGB channels for a spectrogram CNN?

Does it make sense to use an Early Stopping Metric like “mae” instaed of “val_loss” for regression problems?

Machine learning methods for multi-dimensional input and output

Data augmentation or weighted loss function for imbalanced classes?

Does class balancing introduce bias?

How to select the best classifier in classification task?

Link between dimensionality of feature vectors and estimation

When we need fine tuning of CNN?

What is the loss function used for CNN?

Could ResNet curve be used for a regression problem, e.g. housing price prediction?

Why is multicollinearity so bad for machine learning models and what can we do about it?

How initializing weights to a certain variance reduce vanishing / exploding gradients?

When training a neural network to determine appropriate number of epochs should you look at validation loss or validation objective loss?

Does it makes sense to perturb images on-the-run while training CNNs?

What influences fluctuations in validation accuracy?

Stopping criteria for stochastic gradient descent?

Normalizations: dividing by maximum

Why CNNs are less prone to overfitting?

Purpose of Nested Cross-Validation?