Introduction to Computer Vision: Week 2, Exercise 2 Update

I got the below reviewer’s comments about my exercise submission: “You have either exceeded the maximum allowed 9 epochs or your accuracy is below the minimum 0.99 expected value Please try again!”
Kindly assist.

Is the maximum epochs a parameter you can set in your code? What is yours set to? What accuracy do you get? If you set it to 9 (or less) what accuracy do you get? I think the point is that model architectures are not all equal in training effectiveness or efficiency. So these constraints force good choices for layers and their parameters (kernel size, stride, etc). If you aren’t getting to 99% in 9 epochs with your current model architecture, try tweaking it.

Hi @olalekan-ayinde !

Nice to see you here.

Just like @ai_curious mentioned it’s about tweaking your model to try to achieve a certain value of accuracy and this is can be done by using many techniques such as :

  • Changing the learning rate.
  • Adding more layers
  • Using dropout.
  • Using different activation functions.
  • Regularization, etc.

But in this exercise, the main goal is trying to find the best model architecture, since like it’s mentioned you are restricted to a specific number of epochs which is, by the way, a Hyperparameter that we need to tweak and this is done by experimenting different values and see the results so there is no golden rule here.

Hope that helps