fchollet/deep-learning-with-python-notebooks

6.1.3 Why did I get worse when I increased the sample size?

fangxuehouwuming opened this issue · 4 comments

fangxuehouwuming commented

Following the example in the book, I increased the sample size from 200 to 3000 and added a dropout regularization, but the final validation accuracy is only 50%.
1661074009976

ifond commented
lxl0506 commented
fangxuehouwuming commented

My setting is training samples : validation samples =3000 : 10000 . Is it reasonable?
May be I know.

Shuyib commented

@fangxuehouwuming Maybe try to make a train,validation and test set. And see what happens?