How to improve validation error

Part 1 (2019)
1

Hello @jeremy,
Following lesson 1, I have trained a classifier to identify car models from a dataset I found from the link below.
https://ai.stanford.edu/~jkrause/cars/car_dataset.html

I have been trying to improve validation loss, but it doesn’t seem to go down as fast as training loss. Please see the image below.
Looking at training loss, it feels like I am overfitting, but from yesterday’s(lesson 2) notes, you indicated that having validation loss higher doesn’t mean overfitting.
I am kindof confused as to what the next step should be.

Capture
Capture1

source code is at

iyersathya/course-v3/blob/master/nbs/dl1/cars.ipynb

3

Please read the etiquette guide in the FAQ.

1 Like
4

I see you are referring to

  1. Do not @ people if you are not referencing them for a specific reason that requires the attention of that forum member. Be specially mindful in mentioning Jeremy and Rachel.

Extremely sorry if I offended in any way, next time I will be very careful not to repeat the mistake.

1 Like