Lesson 2 In-Class Discussion ✅



(Dhruv Thakur) #593

I’m getting the following plot after running learn.lr_find(). I’ve chosen the lr range that corresponds to the middle drop. I can see that the valid_loss is beginning to increase on the 6th epoch, but what can be the reason for a constant error_rate? Let me know if more context is required.

(Dhruv Thakur) #594

P.S. This is the case when lr is in range (1e-6,4e-6) too.

(Alok) #595

Updating the fastai library fixed it

(Marc Rostock) #596

answered here :slight_smile:

(Mahesh Khatri) #597

Hope this helps


(Mahesh Khatri) #598

Yes. All non apple and non orange images in training set to be classified as Other.


No it’s not a random direction, but the direction of the gradients. The stochastic refers to the fact we draw batches randomly.

(Mahesh Khatri) #600

Rename then with the most correct labels from a human perspective.

(Mahesh Khatri) #601

Michael, Have you seen this ?

(Mahesh Khatri) #602

As Andrew Ng has said, it is like worrying about over population on Mars.

(Jeremy Howard (Admin)) #603

I removed that one since it’s not the approach we recommend.

(Jeremy Howard (Admin)) #604

Please read the etiquette guide in the FAQ.

(José Fernández Portal) #605

I’m facing the same issue and agree that replacing valid_ds with train_ds is not ok. Looking at the code of the ClassificationInterpretation class, it seems that it only works with the validation set. I guess would be great that it would receive a parameter to select the dataset.

(Mahesh Khatri) #606

This may help.


(Jeremy Howard (Admin)) #607

You will get lower error when using higher res images.

Image sizes for different resnets
(Brian Muhia) #608

That’s OK. Autoencoders are too noisy anyway, which means the NN learns the quirks of the autoencoder, rather than what really makes an image a member of a class.

(Mahesh Khatri) #609

In the video, Jeremy has explained this very clearly


Thanks I’ll be sure to check it out. I am using the same dataset so this could be really helpful!

(Mahesh Khatri) #611

@miwojc has kindly shared this Jeremy paper on another forum. It may help.