Agentic AI Master Class : LLM, RAG, AI Agents, XAI

Measure weight daily to drive action and improve outcomes, applying the 'what gets measured gets improved' principle to machine learning with independent and dependent variables, regression, classification, and learning types.

Identify three examples for classification and regression, define clear dependent and independent variables x and y, and learn how to measure and collect data for machine learning.

Practice a Python refresher covering arrays, indexing, shapes, and vectors, then explore numpy and pandas tasks from loading CSVs in Colab to data frames and basic functions.

Explore simple linear regression with a single x and y using a scatter plot and line of best fit. Use r squared, correlation r, and mean absolute error.

Split data into training and test sets, reshape X for scikit-learn, train a linear regression model, and assess performance with r-squared and MAE while covering preprocessing and overfitting concepts.

Discover how linear algebra underpins simple linear regression and how scikit-learn hides the math, from solving for x and y with matrix inverse, determinant, and dot product using numpy.

Learn multiple linear regression with multiple predictors, assess model fit using r-squared and MAE, interpret coefficients, and apply to insurance charges and advertising spend, including outlier handling.

Learn logistic regression and exploratory data analysis to improve classification accuracy, including encoding, handling nulls, train-test split, and evaluating with a confusion matrix and multicollinearity checks using variance inflation factor.

Apply gradient descent to logistic regression, updating intercept and coefficients via the learning rate to minimize binary cross-entropy, converging to optimal parameters.

Examine how independent variables must be independent in logistic regression, address normality and the law of large numbers, and detect and address multicollinearity with VIF using a weather data example.

Explore decision trees with Gini index, and compare bagging and boosting using random forest and XGBoost, while learning when to use accuracy, precision, recall, and F1.

Explore principal component analysis to reduce dimensionality, standardize data, and transform variables into uncorrelated principal components, then apply PCA to an insurance fraud dataset and evaluate model performance.

Learn unsupervised learning for clustering and anomaly detection when labels are absent, using k-nearest neighbors, the elbow method, and K-means for marketing and fraud detection, including customer segmentation.

Understand tensors from scalar to matrix dimensions and how graphs map data flows and operation order inside TensorFlow, including directed acyclic graphs used in complex convolutional networks.

Explore tensor definitions and constants, perform arithmetic operations, indexing, reshaping, and reductions; learn broadcasting, matrix multiplication with different shapes, and eigenvalues and eigenvectors.

Explore how neural networks learn by assigning weights to factors, multiplying inputs, and using activation functions to predict outcomes. Train the model through backpropagation and gradient descent to minimize loss.

Examine how activation functions introduce non-linearity in neural networks, with sigmoid, relu, leaky relu, tanh, softmax, and elu, and their roles in training and multiclass classification.

Explore convex optimization and gradient descent to minimize convex functions, tune model weights through multiple iterations, and understand the roles of cost and loss in machine learning.

Develop a neural network to predict churn and compare accuracy across activation functions (relu, sigmoid, tanh, softmax) using standardization, encoding, and normalization in a TensorFlow sequential model.