Implement MLP Backprop in NumPy | Dataford Interview Questions

Describe how you would implement a 5-node MLP with backpropagation in NumPy. Implement train_mlp(X, y, epochs, learning_rate, seed) using one hidden layer with exactly five tanh units and one sigmoid output. X is a list of numeric rows, y is a list of binary labels, and the function must return predicted labels as a list of integers after full-batch gradient descent. Initialize parameters deterministically from seed, use binary cross-entropy gradients, and make the implementation work for any positive number of input features.

Constraints

X is a non-empty rectangular list of floating-point or integer rows
1 <= len(X) <= 1000
1 <= len(X[0]) <= 100
len(y) == len(X)
Each y value is either 0 or 1
0 <= epochs <= 10000
learning_rate > 0
The hidden layer contains exactly 5 tanh units

Function Signature

def train_mlp(X, y, epochs, learning_rate, seed):

Your question is Implement MLP Backprop in NumPy. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Constraints

X is a non-empty rectangular list of floating-point or integer rows
1 <= len(X) <= 1000
1 <= len(X[0]) <= 100
len(y) == len(X)
Each y value is either 0 or 1
0 <= epochs <= 10000
learning_rate > 0
The hidden layer contains exactly 5 tanh units

Function Signature

def train_mlp(X, y, epochs, learning_rate, seed):

Your question is Implement MLP Backprop in NumPy. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.