see details 1. Clear the session and load the CIFAR10 data into a variable called cifar. (5 points)
2. Create a small training dataset using the first 1000 training images (and the corresponding labels) from CIFAR10. Similarly, create a small test dataset (and the corresponding labels) using the first test 500 images from CIFAR10. (5 points).
3. Create one-hot encoding for the labels for both train and test labels. (5 points)
4. Instantiate a VGG16 convolutional base without the top layer. (5 points)
5. Extract features from the CIFAR10 images so as to fit the conv_base. (40 points)
6. Flatten the features in order to feed them to a densely connected classifier. (5 points)
7. Build a model with one dense layer with 256 units and “relu” activation, one dropout alyer with 50% dropout rate, and a dense output layer with appropriate parameters. (15 points)
8. Compile the model with categorical_crossentropy as the loss function and optimizer_rmsprop with 0.01% learning rate (lr=0.0001). (5 points)
9. Fit the model using 30 epochs. Plot the loss and accuracies. (5 points)
10. Note that the model is likely to have low accuracy. Explain why. (10 points)
Deadline- 1 day
Also, add the comments in the code based on Question Number.