artificial intelligence fall 2014
This is the work from my artificial intelligence class at columbia! There were four assignments, all in python.
hw1: Robotic path planning. Built a search agent comparing different search algorithms and evaluating their performances. Agent finds a path from start to goal of a bounded mxn arena, moving one square at a time. Implemented in breadth first search, depth first search, and A* search.
hw2: Adversarial search. Created human and computer AI's for the game 2048! Different from the original game in that the computer tries to place tiles to make the game the most difficult for the player. Used minimax algorithm with alpha beta pruning for decision making in the AI's. The highest tile it's hit has been 512!
hw3: Machine Learning. Linear regression with one feature and multiple features, gradient descent, plotting the risk function for different learning rates, used the Normal equation. Used support vector machines with different kernels and cross validation to make a classifier. Matplotlib and sklearn to plot.
hw4: Constraint satisfaction problems. Implemented AC-3 and backtracking to solve sudoku puzzles! AC-3 on its own could solve simple puzzles, while backtracking combined with AC-3 could solve all 500 we were given. Backtracking implemented using the minimum remaining value heuristic. Applied forward checking to reduce variables' domains.
I learned so much in this class! These were great assignments. My favorite was hw4, but hw2 was also really fun.