Build a Neural Network
28.9% acceptance
HardDeep Learning
Commonly asked at:
GoogleWhy this matters:
Tests deep understanding of backpropagation – the most important concept for any deep learning role.
Build a neural network with configurable hidden layers, implementing both forward and backward propagation.
Examples
Constraints
- Use ReLU: g(z) = max(0, z) or sigmoid activations
- Implement backpropagation correctly
- Support configurable architecture [n_in, n_1, n_2, ..., n_out]
Tags
transformersdiffusionrlhfdistributed-training