Part (f): During the training process Bob observes that training loss varies widely from one epoch to the next but does not decrease overall and the model does not learn at all. What do you think could be the problem? Justify your answer.

Part (e): For each confusion matrix, compute the following quantities: Accuracy False positive rate False negative rate

The following are multi-part problems.

If a problem is a linearly separable what kind of model should you use? Would you use a deep neural network?

[Bonus] part (i): Suppose we now wanted to implement (binary) linear SVM as a neural network. What loss function should we use?

Part (d): What procedure/algorithm would you use to train the neural network you described in part (c)?

Part (a): What type of learning task is Bob trying to solve? (Be specific. Don’t say something generic like ‘supervised learning’.)

Give two possible causes for a ML model being unfair.

Part (b): Consider a single neuron/unit with weight vector , bias , and activation function . The input is a vector . (If the math does not display: weights w, bias b, activation function f, input x.) What is the output of the neuron/unit as a function of the parameters and input?

How can attention be used for interpretability?