Explain Logistic Regression, Backprop, and Adam
Company: LinkedIn
Role: Data Scientist
Category: Machine Learning
Difficulty: medium
Interview Round: Onsite
Answer the following machine learning fundamentals questions:
1. **Logistic regression**
- Explain how logistic regression works for binary classification.
- Write the model, the sigmoid function, and the binary cross-entropy loss.
- Derive the gradient descent update for the model parameters.
2. **From logistic regression to neural networks**
- Show how logistic regression can be viewed as a one-layer neural network.
- For a simple feedforward neural network, write the forward pass mathematically and derive the backpropagation equations for the weights and biases.
3. **Mini-batch gradient descent**
- Write pseudocode for training a model using mini-batch gradient descent.
4. **Adam optimizer**
- Describe the main components of Adam.
- Explain why Adam scales updates by the square root of the second moment of the gradient.
Quick Answer: This question evaluates understanding of supervised learning and optimization concepts, specifically logistic regression, sigmoid activation and binary cross-entropy, the interpretation of logistic regression as a one-layer neural network, backpropagation derivations, mini-batch gradient descent, and the Adam optimizer.