Phone

+919997782184

Email

support@roboticswithpython.com

Geeks of Coding

Join us on Telegram

Home Forums Assignment NPTEL_ Practical Machine learning with tensor flow Practical Machine Learning with Tensorflow – Assignment 5

Viewing 0 reply threads
  • Author
    Posts
    • #375
      Abhishek TyagiAbhishek Tyagi
      Keymaster

      1)- Your first task is to perform image classification using convolutional neural networks on the CIFAR10 dataset. Please visit ​ this notebook for answering the questions 1 to 5. You are advised to run the notebook with GPU as the hardware accelerator. You can change this setting in the Edit > Notebook settings. Build a convolutional neural network with the architecture mentioned in the notebook and train the model for 10 epochs. What is the range of test accuracy of the model?

      Answer- 65-70

       

       

      2)- Remove the max pooling layers from the above architecture and re-train the model for 10 epochs. Does the test accuracy get better or worse?

      Answer-  Worse

       

       

      3)- After removing the pooling layer in the architecture of the model, how did the number of parameters of the model change? Try and think of why this happened.

      Answer- None of the above (Reduced by ​ 1128448 / 94496 or increased by 1127448 / 94496)

       

       

      4)- We have now trained a convolutional neural network. But let us check if we intuitively understand how filters work. What does the following filter do?
      Filter = [[1/9, 1/9, 1/9], [1/9, 1/9, 1/9], [1/9, 1/9, 1/9]]

      Answer- Blur the Image

       

       

      5)- Your second task is to perform flower classification using transfer learning taught this week. There are two parts: (1) pretrained base mode
      (2) Fine-tuning the model. Please visit <b>​ this​ </b>notebook for answering the following questions. You need to understand the code and try to fill up the empty cells in the notebook.
      What is the shape of one batch of image_data?

      Answer- [30, 160, 160, 3]

       

       

      6)- What is the shape of a new block of features converted by the feature extractor for each image?

      Answer- [5, 5, 512]

       

       

      7)- What is the total number of trainable ​ <i>TF</i> variables ​ in the model (part-1)?

      Answer- 2

       

       

      8)- What is the total number of trainable parameters in this model (part-1)? Select the order of it:

      Answer- 1000

       

       

      9)- What is the total number trainable <i>​ TF</i> variables ​ in the model (part-2)?

      Answer- 4

       

       

      10)- What is the total number of trainable parameters in this mode (part-2)? Select the order of it:

      Answer- 1000000

       

       

      11)- Let the validation accuracy in part-1 be ​ <i><b>x</b></i> ​ and validation accuracy in part-2 be ​ y ​ (both after training is done)​ . ​ The value |y-x| lies between?

      Answer- 0.0 to 0.1

Viewing 0 reply threads
  • You must be logged in to reply to this topic.