Artificial Intelligence & Machine Learning interview questions and answers

AI is the broad field of building intelligent systems. ML is a subset of AI where systems learn from data. Deep learning is a subset of ML that uses multi-layer neural networks.

Gradient descent is an optimisation algorithm that iteratively adjusts model parameters in the direction that reduces the loss function.

A CNN is a neural network designed for grid-like data such as images. It uses convolutional layers to automatically learn spatial features like edges and shapes.

Regularisation adds a penalty to the loss function — such as L1 or L2 — to discourage overly complex models and reduce overfitting. Dropout is a common technique in neural networks.

Precision is the share of predicted positives that are correct. Recall is the share of actual positives that the model identifies. The F1 score balances both.