Description
Programming Assignment
TOPIC: ARTIFICIAL NEURAL NETWORK
1. Build a Convolutional Neural Network for handwritten digit classification.
a. Print the model summary – layers used in the model and the specifications
b. Display the validation accuracy after each epoch
Specifications:
– Dataset: MNIST
– Batch size: 32
– Epochs: 30
– Optimizer: Adam
– Loss: Cross Entropy Loss
– Activation function: tanh
2. Add pooling layers to the model, print model summary, and verify the number of parameters and
accuracy of the model
3. Use ReLU instead of tanh, recalculate the accuracy, and write the observations.
4. Build a standard Artificial Neural Network (without any convolution layers) for handwritten digit
classification.
a. Print the model summary
b. Display the validation accuracy on each epoch
Instructions for submission
1. For each question, there will be one folder containing the code and the results.
2. Zip all such folders into one zipped folder named as <Roll No>AssignmentANN.
