User Ben Reiniger

155 votes

19 answers

185k views

How do you visualize neural network architectures?

Martin Thoma

17.6k

asked Jul 18, 2016 at 17:08

59 votes

5 answers

24k views

In softmax classifier, why use exp function to do normalization?

machine-learning deep-learning

Hans

693

asked Sep 20, 2017 at 5:53

55 votes

6 answers

15k views

Should I go for a 'balanced' dataset or a 'representative' dataset?

machine-learning dataset class-imbalance

pnp

653

asked Jul 22, 2014 at 12:29

45 votes

5 answers

22k views

Is it always better to use the whole dataset to train the final model?

machine-learning dataset training accuracy

pcko1

3,740

asked Jun 12, 2018 at 9:54

19 votes

4 answers

43k views

L2 loss vs. mean squared loss

loss-function

Edamame

2,505

asked Jan 1, 2018 at 1:58

15 votes

3 answers

16k views

L1 & L2 Regularization in Light GBM

xgboost regularization lightgbm

Vikrant Arora

376

asked Aug 8, 2019 at 17:08

12 votes

2 answers

4k views

What is the proper way to use early stopping with cross-validation?

xgboost cross-validation lightgbm early-stopping

Amine SOUIKI

121

asked May 17, 2020 at 15:15

11 votes

2 answers

2k views

What is the most efficient method for hyperparameter optimization in scikit-learn?

scikit-learn hyperparameter hyperparameter-tuning grid-search randomized-algorithms

Brian Spiering

16.9k

asked Mar 13, 2019 at 19:42

8 votes

2 answers

104 views

Model for Differing Number of Rows per Observation

predictive-modeling feature-selection feature-engineering

Zachary

181

asked Apr 17, 2019 at 16:47

8 votes

2 answers

199 views

Time-series prediction: Model & data assumptions in AI/ML models vs conventional models

machine-learning neural-network time-series linear-regression

Maeaex1

538

asked Aug 29, 2019 at 6:45

8 votes

2 answers

89 views

Which classification algorithms are negatively affected by class imbalances?

machine-learning classification predictive-modeling multilabel-classification class-imbalance

Danny David Leybzon

180

asked Jul 3, 2019 at 19:45

8 votes

2 answers

2k views

Transformer model: Why are word embeddings scaled before adding positional encodings?

tensorflow nlp transformer attention-mechanism

Milad Shahidi

333

asked Jan 13, 2021 at 10:10

7 votes

2 answers

305 views

Dropping features after final evaluation on test data

machine-learning feature-selection random-forest overfitting

Alireza Amani

123

asked Dec 29, 2020 at 17:37

7 votes

1 answer

551 views

Reticulate vs Python

python r

Dave

2,915

asked Aug 7, 2019 at 1:47

7 votes

4 answers

2k views

Can the vanishing gradient problem be solved by multiplying the input of tanh with a coefficient?

machine-learning neural-network deep-learning activation-function

zephyr

121

asked May 7, 2019 at 13:07

6 votes

1 answer

480 views

Strategies for handling unlabeled data which is slightly different from the labeled data

machine-learning python classification

Mathias

69

asked Apr 4, 2019 at 9:37

6 votes

1 answer

863 views

Understanding LSTM structure

deep-learning keras lstm rnn

kg_sYy

161

asked Feb 18, 2019 at 22:20

6 votes

1 answer

2k views

Smart data split (train/eval) for Object Detection

dataset data training evaluation object-detection

CarlosUziel

61

asked Jun 25, 2019 at 11:44

6 votes

2 answers

207 views

ground truth fit is worse than cross validated fit on noisy data?

python regression cross-validation

user2183045

69

asked Jul 6, 2020 at 8:52

5 votes

3 answers

180 views

How to use "tree boosting" with a data-driven loss function

machine-learning random-forest decision-trees xgboost loss-function

David

51

asked Oct 2, 2020 at 18:25

5 votes

1 answer

5k views

svm.LinearSVC: larger max_iter number doesn't always increase the accuracy/precision/recall

scikit-learn svm supervised-learning model-selection

TJCLK

281

asked Aug 22, 2019 at 8:29

5 votes

1 answer

1k views

Multiclass classification on imbalanced dataset : Accuracy or micro F1 or macro F1

multiclass-classification accuracy

Bikash Gyawali

151

asked May 11, 2019 at 22:40

5 votes

2 answers

117 views

Does a precision score increasing with a higher number of folds mean the model will improve with more data?

classification nlp cross-validation evaluation

Robert Patchouli

51

asked Feb 13, 2019 at 18:02

5 votes

3 answers

933 views

Quasi-categorical variables - any ideas?

time-series regression categorical-data

Uri Merhav

206

asked Jan 28, 2015 at 13:42

5 votes

1 answer

2k views

My CNN produces volatile validation_loss and does not converge in python(Keras)

python keras cnn

BloodRabz

51

asked May 7, 2019 at 10:34

5 votes

1 answer

213 views

Feature scaling worsens performance?

python neural-network implementation

Soumirai

151

asked May 9, 2019 at 16:07

4 votes

1 answer

1k views

AttributeError: type object 'Lambda' has no attribute 'shape'

python keras tensorflow

Kahina

574

asked May 2, 2019 at 17:25

4 votes

3 answers

444 views

Train classifier on balanced dataset and apply on imbalanced dataset?

scikit-learn data class-imbalance gbm

user3240855

41

asked Mar 5, 2019 at 16:10

4 votes

1 answer

281 views

Hyperopt vs Default Values

classification random-forest hyperparameter

Sarah

571

asked May 27, 2019 at 23:52

4 votes

2 answers

3k views

Why does feature scaling improve the convergence speed for gradient descent?

linear-regression gradient-descent supervised-learning

Yandle

141

asked Jul 14, 2019 at 19:33

Stack Exchange Network

current community

your communities

more stack exchange communities

91 Bookmarks

How do you visualize neural network architectures?

In softmax classifier, why use exp function to do normalization?

Should I go for a 'balanced' dataset or a 'representative' dataset?

Is it always better to use the whole dataset to train the final model?

L2 loss vs. mean squared loss

L1 & L2 Regularization in Light GBM

What is the proper way to use early stopping with cross-validation?

What is the most efficient method for hyperparameter optimization in scikit-learn?

Model for Differing Number of Rows per Observation

Time-series prediction: Model & data assumptions in AI/ML models vs conventional models

Which classification algorithms are negatively affected by class imbalances?

Transformer model: Why are word embeddings scaled before adding positional encodings?

Dropping features after final evaluation on test data

Reticulate vs Python

Can the vanishing gradient problem be solved by multiplying the input of tanh with a coefficient?

Strategies for handling unlabeled data which is slightly different from the labeled data

Understanding LSTM structure

Smart data split (train/eval) for Object Detection

ground truth fit is worse than cross validated fit on noisy data?

How to use "tree boosting" with a data-driven loss function

svm.LinearSVC: larger max_iter number doesn't always increase the accuracy/precision/recall

Multiclass classification on imbalanced dataset : Accuracy or micro F1 or macro F1

Does a precision score increasing with a higher number of folds mean the model will improve with more data?

Quasi-categorical variables - any ideas?

My CNN produces volatile validation_loss and does not converge in python(Keras)

Feature scaling worsens performance?

AttributeError: type object 'Lambda' has no attribute 'shape'

Train classifier on balanced dataset and apply on imbalanced dataset?

Hyperopt vs Default Values

Why does feature scaling improve the convergence speed for gradient descent?