AI Engineer Interview Questions and Answers

Key assumptions of linear regression include linearity, independence, homoscedasticity, normality, no multicollinearity, and no endogeneity. Techniques for handling imbalanced data in classification include resampling, adjusting class weights, generating synthetic data, and using appropriate metrics like F1-score. Backpropagation in neural networks involves a forward pass, loss calculation, backward pass to compute gradients, and weight updates using optimization techniques. The Transformer architecture uses self-attention mechanisms and has advantages like parallel processing and capturing long-range dependencies.