Essentials of Machine Learning

Explore the essentials of machine learning, including key terms, data preparation, performance measures, regression and classification algorithms, and optimizers, with insights into data-to-production workflows.

Cover supervised, unsupervised, and reinforcement approaches, including regression and dimensionality reduction. Learn about training, validation, testing, hyper parameters, grid search, and cross-validation to build and evaluate models while avoiding overfitting.

Explore how models approximate reality, compare several candidates, and learn key terms like features, observations, X, y, bias, weights, and the parameter vector.

Contrast batch learning and online learning, then emphasize data understanding and preparation, bias and variance, regularization, and train-validation-test evaluation with retraining to stay current.

Learn data preparation techniques to handle skewed distributions, apply log scaling or a threshold, use percentiles and miniluxe scaling, and standardize features for consistent input ranges.

Learn data preparation techniques, including sampling bias and stratified sampling to mirror training and testing distributions. Oversample rare events when needed, and encode categorical data with one-hot encoding or embeddings.

Explore classification algorithms that assign data to predefined categories or groups, contrast supervised and unsupervised learning, and compare logistic regression and linear support vector machine using the iris dataset.

Explore multiclass supervised algorithms, including one-vs-rest logistic regression and k-nearest neighbors, and cluster data with k-means, while evaluating with accuracy, precision, recall, and confusion matrices.

Explore regression algorithms that predict numerical values, fill in missing data, and forecast outcomes, from linear regression to polynomial fits, with evaluation via mean absolute error.

Explore optimization techniques to speed up learning in machine learning models, including gradient descent, momentum, Nesterov Accelerated Gradient (Nag), RMSProp, and Adam, and tune the learning rate for balance.