Introduction to AI - Machine Learning and Deep Learning

Refresh statistics basics—mean, median, mode, variance, standard deviation, and outliers—then cover probability distributions, correlation, Bayes’ theorem, calculus for slopes, and matrix algebra.

Explore key function types for machine learning, including linear, quadratic, cubic, and exponential and logarithmic functions, with notes on slope, differentiability, and leaky ReLU lab implementation.

Explore a quick lab on mathematical functions, plotting y=f(x) from simple linear to nonlinear forms, including relu and leaky relu, to sharpen skills for machine learning algorithms.

Learn core data concepts for machine learning, including discrete, continuous, categorical, and time series data, and master data preparation, imputation, normalization, adding or removing features, one-hot encoding, SMOTE, and transformations.

Explore machine learning terminology including observation, attribute, and target attribute, and explain training, validation, and test sets, loss and cost, forward and backward propagation, gradient descent, learning rate, and hyperparameters.

Explore core machine learning and deep learning algorithms, including supervised, unsupervised, semi-supervised, and reinforcement learning, with practical examples and intuitive explanations.

Explore regression in machine learning, including linear and logistic regression, ordinary least squares, and multivariate regression, with practical code using NumPy and scikit-learn.

Learn to perform linear regression with sklearn in a Jupyter notebook by modeling y as 2x1 plus 5x2 plus 10, fitting the model, and achieving a perfect score with predictions.

Explore logistic regression with sklearn, modeling binary outcomes using a four by two matrix and a one by four y vector, fit the model, and interpret coefficients and intercept.

Perform multivariate regression with feature engineering and ordinary least squares on house price data, using rooms, area, and L1, L2, L3, scaled to 0–1, and predict a $600,000 price.

Learn how decision trees use maximum information gain to split binary and numeric features for multi-class classification, and how ensemble methods like bagging and boosting (XGBoost) improve performance.

Show how to use sklearn to build a decision tree on seven-row animal and bird data set with one-hot targets, fit and predict in Jupyter Notebook, and visualize the tree.

Train a decision tree on the iris dataset with sklearn, fit the model and predict flower types, shown by a plot that reveals how decisions are made and predicts virginica.

Learn how to apply XGBoost to the iris dataset from Kaggle by importing the XGBoost classifier, splitting the data into train and test, training the model, and predicting iris virginica.

Discover how neural networks work from input to output layers, with weights and activation functions, and explore concepts like CNNs, RNNs, LLMs, GANs, dropout, and softmax.

Train a Keras neural network on the MNIST handwritten digits dataset with 60,000 training and 10,000 test samples, using an input layer and two 512-node dense layers over 10 epochs.

Explore the K-means clustering algorithm for unlabeled data, learning how centroids define clusters and how gradient descent optimizes the grouping to minimize intra-cluster distances.

Evaluate model performance after training using RMSE for numeric outputs and confusion matrix metrics: precision, recall, F1, and AUC for classification; apply MAPE for time series.

Explore advanced algorithms from factorization machines and k-nearest neighbors to anomaly detection, word2vec, and sequence2sequence, plus pca, image classification, object detection, and topic modeling with lda.