Deep learning fundamentals: optimization, drift and calibration

Answer conceptual questions on training: gradient descent versus SGD, convexity, sharp versus flat minima, the bias-variance trade-off, calibration and model drift, and how you would frame an ML quality problem for the business.