Skip to content

README.md

Latest commit

 

History

History
61 lines (46 loc) · 5.73 KB

README.md

File metadata and controls

61 lines (46 loc) · 5.73 KB

CS540 Introduction to Artificial Intelligence

This repository contains my coursework for CS540: Introduction to Artificial Intelligence at the University of Wisconsin-Madison, Fall 2024.

Course Information

  • Course Description: Gain principles of knowledge-based search techniques; automatic deduction, knowledge representation using predicate logic, machine learning, probabilistic reasoning. Develop applications in tasks such as problem solving, data mining, game playing, natural language understanding, and robotics.
  • Credits: 3
  • Prerequisites: (COMP SCI 300 or 367) and (MATH 211, 217, 221, or 275) or graduate/professional standing or declared in the Capstone Certificate in Computer Sciences for Professionals.
  • Sections:
    • Section 1: MW 4:00 - 5:15, 145 Birge Hall
    • Section 2: TR 2:30 - 3:45, 19 Ingraham Hall
    • Section 3: TR 1:00 - 2:15, 5206 Sewell Social Sciences
  • Textbook (Optional): Artificial Intelligence: A Modern Approach (4th edition) by Stuart Russell and Peter Norvig. Pearson, 2020. ISBN 978-0134610993.

Course Objectives

  • Understand and apply foundational tools in Machine Learning and Artificial Intelligence: Linear algebra, Probability, Logic, and elements of Statistics.
  • Understand core techniques in Natural Language Processing (NLP), including bag-of-words, tf-idf, n-Gram Models, and Smoothing.
  • Understand the basics of Machine Learning, including supervised and unsupervised learning.
  • Distinguish between regression and classification, and understand basic algorithms: Linear Regression, k-Nearest Neighbors, and Naive Bayes.
  • Understand the basics of Neural Networks: Network Architecture, Training, Backpropagation, Stochastic Gradient Descent.
  • Learn aspects of Deep Learning, including network architectures, convolution, training techniques.
  • Understand the fundamentals of Game Theory.
  • Understand how to formulate and solve several types of Search problems.
  • Understand basic elements of Reinforcement Learning.
  • Consider the applications and ethics of Artificial Intelligence and Machine Learning in real-world settings.

Projects

  • Letter Frequency Analysis: https://github.com/Eddylin03/CS540/blob/main/hw2/hw2.py - This script analyzes the frequency of letters in a given text file and uses Bayesian reasoning to determine if the text is more likely to be in English or Spanish.

  • Facial Recognition with Eigenfaces: https://github.com/Eddylin03/CS540/blob/main/hw3/Demonstration.ipynb - Implemented a facial recognition system using Principal Component Analysis (PCA) to identify individuals from a dataset of facial images.

  • Hierarchical Clustering of Socioeconomic Data: https://github.com/Eddylin03/CS540/blob/main/hw4/hw4.py - Applied hierarchical agglomerative clustering (HAC) to analyze socioeconomic indicators from various countries, visualizing country clusters based on socioeconomic profiles.

  • Linear Regression on Lake Mendota Ice: https://github.com/Eddylin03/CS540/blob/main/hw5/hw5.py - Implemented a linear regression model to analyze historical ice coverage days on Lake Mendota. Tasks included dataset curation, data normalization, closed-form solution calculation, gradient descent optimization, and prediction for future ice coverage based on trends.

  • Neural Network with PyTorch: https://github.com/Eddylin03/CS540/blob/main/hw6/intro_pytorch.py - Developed a basic neural network in PyTorch to classify images from the FashionMNIST dataset. Implemented data loading, model building, training, evaluation, and individual image prediction to familiarize with PyTorch's capabilities.

  • LeNet on MiniPlaces Dataset: https://github.com/Eddylin03/CS540/blob/main/hw7/demo.ipynb - Implemented the LeNet-5 convolutional neural network architecture to perform scene recognition on the MiniPlaces dataset. Built a custom CNN model with two convolutional layers and three fully connected layers, explored various hyperparameter configurations including batch size and learning rate, and analyzed model performance across different training settings.

  • A Search for N-Tile Puzzle:* https://github.com/Eddylin03/CS540/blob/main/hw8/Demo.ipynb - Developed an A* search implementation to solve variants of the 8-tile sliding puzzle with multiple empty spaces. Implemented Manhattan distance heuristic for state evaluation, dynamic successor state generation, and efficient priority queue-based pathfinding. The solution handles various puzzle configurations (6-tile, 7-tile, etc.) and includes comprehensive state validation.

  • [...and so on...]

Technologies Used

  • Python
  • NumPy - For numerical operations, particularly useful for matrix manipulations in linear regression and gradient descent.
  • Matplotlib - Used for visualizing data plots, loss plots, and other figures in assignments.
  • PyTorch - Utilized for building, training, and evaluating neural networks in deep learning assignments.
  • Pandas - Employed for data handling and cleaning, especially in preprocessing datasets.
  • Torchvision - Part of PyTorch, used to load and transform image data (like FashionMNIST) for neural network training.
  • heapq - Python's built-in heap queue algorithm implementation, used for priority queue in A* search puzzle solving.
  • tqdm - Progress bar library for monitoring training progress in neural network implementations.
  • shadcn/ui - React component library utilized for building modern user interfaces in web-based demonstrations.
  • torch.nn - PyTorch's neural network modules, essential for implementing LeNet architecture.
  • torch.optim - PyTorch's optimization algorithms for training neural networks with different learning rates.

Notes

  • This repository will be updated throughout the semester as I complete assignments and projects.
  • Feel free to reach out if you have any questions!