using test set as validation set
bz36912 opened this issue · comments
Issue Type
Bug
Source
source
Keras Version
Keras2.13
Custom Code
Yes
OS Platform and Distribution
Windows 11
Python version
3.8.5
GPU model and memory
No response
Current Behavior?
In examples/vision/pointnet.py. You seem to use used your test set as validation set in line 263:
model.fit(train_dataset, epochs=20, validation_data=test_dataset)
This would cause the performance and predictive ability of your model to be overly optimistic, since the test set is not separate from your training process. The model is likely to perform poorly with unseen data/points.
Standalone code to reproduce the issue or tutorial link
examples/vision/pointnet.py line 263:
`model.fit(train_dataset, epochs=20, validation_data=test_dataset)`
Relevant log output
No response
I'd like to contribute. @sachinprasadhs