| Applying Math and Probability concepts (k3): Students can apply fundamental concepts of math and especially
probability theory to machine learning methods, including Gaussian distributions, expectations, covariances,
etc.
Solving Optimization Problems (k5): Students can formulate and solve optimization problems for machine
learning models in closed form or using iterative algorithms, such as various variations of gradient descent.
Implementing Supervised Learning Models (k5): Students can code and analyze models for regression and
classification as well as understand their hyperparameters.
Evaluating Predictors (k5): Students can evaluate model performance by analyzing various error types as well
as knowing about the bias and variance tradeoff.
Deriving MLE and MAP Estimators (k5): Students can estimate model parameters using Maximum Likelihood
Estimation (MLE) and Maximum A Posteriori (MAP) frameworks.
Understand and implement basic RL concepts (k4): Students can formalize and implement problems as Markov
Decision Processes, know the basic elements of reinforcement algorithms, and understand basic exploration
methods as well as the exploration/exploitation tradeoff.
|
Students acquire a comprehensive understanding of the mathematical foundations of machine learning,
encompassing empirical risk minimization, loss functions, and probabilistic frameworks such as maximum likelihood
and maximum a posteriori estimation. They understand the theoretical connection between approximation,
estimation, and irreducible errors, specifically within the context of the bias-variance decomposition. Students will
also learn about Markov decision processes, value functions, and exploration in the reinforcement learning context.
Furthermore, they will possess mathematical tools from probability theory and optimization (such as gradient
descent) that serve as the computational engines for the vast majority of learning approaches.
|
