Machine Learning: Making computers think!

Learn descriptive statistics to summarize data with mean, median, mode, and dispersion measures like standard deviation and variance, and understand how inferential statistics draw conclusions about populations from samples.

Master data preprocessing as a prelude to modeling by transforming incomplete raw data into an understandable format, using standardization, normalization, and handling categorical data and missing values.

Explore practical techniques for handling categorical data by encoding methods such as dummy variables and one-hot or label encoding, turning characters into numbers for model-ready input.

Classify data with tree-based and non-tree-based models, exploring decision trees, Bayes theorem, nearest neighbors, linear regression, and k-means clustering, while balancing interpretability and accuracy.

Prepare the iris data by creating train, validation, and test sets before modeling. Train a baseline model, then use SVM and k-means to generate values and visualize clusters with centroids.

Explore tuning and validation concepts, learn about validation metrics, and improve accuracy through 14 topics, culminating in hyper barometer tuning.

Explore two validation approaches for machine learning: a train/validation split with 20% validation yielding 0.967 accuracy, and cross-validation that estimates model performance across data subsets.

Balance bias and variance to improve generalization, distinguishing overfitting from underfitting as models trade training accuracy for test accuracy. Understand how mapping and target functions shape train versus test performance.

Explore methods to improve machine learning accuracy by adding data, engineering features, selecting the best features, combining multiple weak learners through sampling, and tuning models for stronger performance.

Explore hyperparameter tuning to improve accuracy by adjusting learning rate and gradient descent parameters, using SVM as a baseline and selecting parameter values based on training data.

Explore ensemble learning techniques, why sampling matters, and a look at bagging and boosting to improve accuracy.

Explore bagging (bootstrap aggregating) and boosting, building strong learners from multiple weak models, and explain stacking with a meta-model combining base predictions for higher accuracy.

Demonstrate the curse of dimensionality by showing how increasing dimensions hinder finding a target, and explain dimensionality reduction as a way to preserve information for classification or regression.