Background
You are tasked with classifying images of clothing from the Fashion MNIST dataset using a Neural Network model built with Tensorflow and Keras. The dataset includes 60,000 training images and 10,000 test images, each of size 28x28 pixels, representing different classes of fashion items.
Questions:
1. Data Loading and Exploration
● What dataset is used in this case study, and how is it loaded into the model?
● What are the dimensions of the training and test images, and what do the labels represent?
● How many classes are there, and what are the names of these classes?
2. Data Preprocessing
● Why is it important to scale the pixel values between 0 and 1?
● What function or method is used to perform this scaling?
3. Model Building and Compilation
● What type of model architecture is used in this solution? Briefly describe its layers.
● Which activation functions are used in the model, and why?
● What loss function and optimizer are chosen for this model? Explain why they are suitable for this task.
4. Model Training
● Explain the concept of an epoch in model training. How many epochs were used in this case study?
● What does the batch size parameter control, and how does it affect model training?
5. Evaluation
● What metrics are used to evaluate the model’s performance?
● What was the final test accuracy achieved by the model, and how does it compare with training accuracy?
6. Predictions and Visualizations
● How does the model make predictions, and what method is used to identify the predicted class?
● What visualization techniques are used to understand the model's predictions? Explain the color coding for correct and incorrect predictions.