Explore mathematical optimization techniques for AI problem-solving in this comprehensive conference talk. Dive into constraint-based optimization, feasibility vs. optimization, and various problem classes. Learn to define optimization models using decision variables, constraints, and objectives. Examine real-world examples like Sudoku and scheduling problems. Gain practical insights into implementing these concepts using Google OR-Tools, with a focus on code implementation rather than complex mathematics. Discover how to efficiently solve problems with multiple solutions in AI systems, including capacity utilization, shortest path finding, and optimal scheduling. Perfect for software developers looking to enhance their AI development skills with optimization techniques.
Overview
Syllabus
Intro
What do I mean by "Artificial Intelligence"?
Types of Al Models
Constraint-Based Optimization
Classes of Problems
Feasibility vs. Optimization
Constraint: A Required Condition
Types of constraints
Objective: A Goal for the Solution
Constraint Programming: Sudoku
Linear Example: Pete's Pottery Paradise
Pottery Production - Solution Space
Pottery Production - The Polytope
Defining an Optimization Model
Decision Variables
Execute the Model
Mixed-Integer Example: Scheduling
LP Model - Variables
LP Model - Constraints
MIP Model - Variables
MIP Model - Constraints
MIP Model - Objective Example
Summary
Resources
Taught by
NDC Conferences
