Foundations of A.I.: Knowledge Representation & Learning

Explore the four AI paradigms—think rationally, think like humans, act optimally, and act like humans—with examples like Google Maps, Siri, and Roomba.

this lecture surveys AI paradigms, including symbolic AI, connectionist AI, artificial general intelligence, cognitive systems, and explains how machine learning learns from data and deep neural networks drive deep learning.

Explore how artificial intelligence applies to autonomous driving, games, and robotics. Examine tools like sensors for perception, path planning, object detection, and decision making under uncertainty.

Explore installing the Anaconda distribution on Windows 10, including verifying system type, downloading the 64-bit graphical installer, and completing the installation steps, with notes on mac or Linux differences.

Navigate the Anaconda Navigator interface and open Jupyter Notebook to begin learning Python. Explore Spyder for scientific programming, and learn how to use the environment tab to manage navigator versions.

Learn two methods to open Jupyter Notebook: use the launch button in Anaconda Navigator or the start menu, then view the Jupyter Notebook home page in your browser.

Transform problem solving agents into knowledge based agents by representing the world internally with a knowledge base and inference engine, which derive decisions from rules and update knowledge.

Explore how humans and machines represent knowledge through syntax, semantics, and inference, building a knowledge base and using logic to derive conclusions.

Explore knowledge representation techniques, including logical representation with propositions, syntax and semantics, semantic networks, frames, and production rule systems, with emphasis on semantic networks and expert systems.

Explore expert systems that imitate expert reasoning by combining a knowledge base of facts and heuristic rules with an inference engine to handle uncertainty in health care and aviation.

Build an expert system in Python using the knowledge engine and exporter library, defining facts, rules, and diagnoses for Covid 19 symptoms in basic and advanced modes.

Explore semantic networks as graph-based knowledge representations using nodes and arcs to express binary relationships, inheritance, part-of connections, has, and causal links.

Construct a semantic network with network X to represent Mark's statements as a directed, weighted graph of nodes and edges.

Explore how learning agents adapt through experience, using sensors, actuators, a performance element, a critic, and a learning element. Mitchell defines learning as performance on a task improving with experience.

Learn how machine learning enables systems to learn from data, covering supervised, unsupervised, and reinforcement paradigms with labeled data, unlabeled data, and reward-based trial and error.

Learn how decision trees perform supervised learning by splitting data on attributes, using root and leaf nodes, to classify outcomes and predict decisions with cart.

Build a decision tree classifier in Python with scikit-learn to classify certificates from attendance, theory, and daily assessments. Preprocess data, split into train/test sets, and evaluate with entropy and accuracy.

Decision trees offer an easy, interpretable model for classification and prediction, used in fraud detection, grade classification, healthcare decisions, CRM insights, and energy consumption analysis.

Explore simple linear regression, modeling the relationship between a single input x and output y, using the regression line, slope, and y-intercept to predict future values.

Implement a simple linear regression in python using scikit-learn to predict tips from total bill, processing the tips dataset, splitting data into train and test, fitting, predicting, and visualizing results.

Apply linear regression to forecast sales, inventory, stock prices, and real estate values using past data and features such as opening price, closing price, highest price, lowest price, and volume.

Explore how artificial intelligence relies on knowledge representation and learning, from knowledge graphs and semantic networks to learning agents and decision trees for real-world predictions.