121 - 130 of 236 results for: MS

Markets are everywhere around us but don't always achieve desired goals. Market failures occur due to a variety of frictions and need design to be fixed. The design of marketplace varies depending on the type of goods and possible transactions. This course will cover methods and classic results to analyze the behavior of a marketplace, whether it is successful and how to fix it, building especially on game theoretic tools. The course will further explore the trade-offs between efficiency and equitable outcomes and how to reach desired outcomes. Applications include matching students to schools, college admissions and the failure the desire to balance equity and merit, assigning vaccines, assigning interns to hospitals, assigning organs to patients, auction designs and pricing, information design, online platforms, allocation of food, transportation, and emissions. The course is intended for undergraduates, masters, but also PhD students who are interested in exposure to market design. Prerequisites: basic mathematical maturity at the level of Math 51, and probability at the level of MS&E 120, 220 or EE 178. Limited enrollment.

Examines foundations of strategic environments with a focus on game theoretic analysis. Provides a solid background to game theory as well as topics in behavioral game theory and the design of marketplaces. Introduction to analytic tools to model and analyze strategic interactions as well as engineer the incentives and rules in marketplaces to obtain desired outcomes. Technical material includes non-cooperative and cooperative games, behavioral game theory, equilibrium analysis, repeated games, social choice, mechanism and auction design, and matching markets. Exposure to a wide range of applications. Lectures, presentations, and discussion. Prerequisites: basic mathematical maturity at the level of Math 51, and probability at the level of MS&E 120 or EE 178.

Game theory uses mathematical models to study strategic interactions and situations of conflict and cooperation between rational decision-makers. This course provides an accelerated introduction to tools, models and computation in non-cooperative and cooperative game theory. Technical material includes normal and extensive form games, zero-sum games, Nash equilibrium and other solution concepts, repeated games, games with incomplete information, auctions and mechanism design, the core, and Shapley value. Exploration of applications of this material through playing stylized in-class and class-wide games and analyzing real-life applications. Prerequisites: mathematical maturity at the level of MATH51, and probability at the level of MS&E 120, or equivalent.

The course explores applied topics at the intersection of game theory, data science, and artificial intelligence. The first part of the course focuses on computational approaches to solving complex games (such as Poker), with applications in developing successful algorithmic agents for playing these games. Lectures provide the foundations of the methods underlying these computational game theory approaches (rooted in the theory of learning in games and online learning theory) and proofs of their properties, while assignments involve implementing simple variants. The second part of the course examines the interplay between data science and mechanism design. Topics include optimizing auctions and mechanisms from data, with applications to online auction markets. The course also covers methodologies for learning structural parameters in games and econometrics in games, and how these techniques are used to analyze data from strategic interactions, such as auction data. The third part of the course explores topics related to A/B testing in markets with strategic interactions, with applications to digital matching platforms such as Airbnb. Prerequisites include mathematical maturity in probability, statistics, optimization, linear algebra, and calculus.

Formulation and solution of sequential decision problems under uncertainty as a foundation for artificial intelligence, operations research, and economics. Finite-horizon, discounted, and average reward objectives. Optimization via value iteration, policy iteration, linear programming, and reinforcement learning algorithms. Semi-Markov decision processes. Multi-armed bandits and the Gittin's index theorem. Use of partial state information. Homework assignments involve a combination of analytic and computational exercises.

The subject of reinforcement learning addresses the design of agents that improve decisions over time while operating within complex environments. This course covers desired agent behaviors and principled scalable approaches to realizing such behavior. Homework assignments primarily involve programming exercises carried out in Colab.

Machine learning has become a powerful tool for discrete optimization. This is because, in practice, we often have ample data about the application domain?data that can be used to optimize algorithmic performance, ranging from runtime to solution quality. This course covers how machine learning can be used within the discrete optimization pipeline from many perspectives, including how to design novel combinatorial algorithms with machine-learned modules and configure existing algorithms? parameters to optimize performance. Topics will include both applied machinery (such as graph neural networks, reinforcement learning, transformers, and LLMs) as well as theoretical tools for providing provable guarantees.

The subject of reinforcement learning addresses the design of agents that improve decisions over time while operating within complex and uncertain environments. This first course of the sequence restricts attention to the special case of bandit learning, which focuses on environments in which all consequences of an action are realized immediately. This course covers desired agent behaviors and principled scalable approaches to realizing such behavior. Topics include learning from trial and error, exploration, contextualization, generalization, and representation learning. Motivating examples will be drawn from recommendation systems, crowdsourcing, education, and generative artificial intelligence. Homework assignments primarily involve programming exercises carried out in Colab, using the python programming language and standard libraries for numerical computation and machine learning. Prerequisites: programming (e.g., CS106B), probability (e.g., MS&E 121, EE 178 or CS 109), machine learning (e.g., EE 104/ CME 107, MS&E 226 or CS 229).

This project-based experiential course is designed for advanced undergraduate and masters students familiar either with market design basics or machine learning methods who are interested in studying and potentially building a platform in a specific application domain. Applications of interest include the sharing economy, online advertising, blockchains and decentralized finance, as well as markets for allocating public goods. Prerequisites: 230, 260, CS 230, or equivalents.