Batch size for Feed-Forward Neural Network

Question

I have a 100k dataset for ML, how much should I set up the batch size for training, ?

FYI - I am using the train_test_split library to splitting the data into train and test sets.

Thank you!

## Fit network

history = model.fit(X_train, Y_train, epochs = 1500, validation_split=0.2, verbose=1, shuffle=False, batch_size=60)

score 0 · Answer 1 · answered Feb 01 '23 at 03:04

0

Wannees, You need to give more details so that others can help you. Such as type of data (numeric or image), application, number of features/ targets, network architecture, etc.

You may learn something from the existing question in this web: How to calculate optimal batch size, https://stats.stackexchange.com/questions/153531/what-is-batch-size-in-neural-network

answered Feb 01 '23 at 03:04

Harry

86
1
5

Harry - thank you so much for your responses. my data are numeric, the number of features is 5 and the target is one. I am using techniquemany to one. – Wannees Alkhayyali Feb 01 '23 at 17:30
Wannees, I suggest to use the batch size in these sets [2, 4, 8, 16, 32, 64]. Can can try 32 first. – Harry Feb 04 '23 at 18:00
Will do. thank you so much, Harry! – Wannees Alkhayyali Feb 04 '23 at 20:20
Keep posting, I can follow this topic – Harry Feb 05 '23 at 02:02

Batch size for Feed-Forward Neural Network

1 Answers1