Randomized Rounding Approaches to Online Allocation, Sequencing, and Matching

Will Ma

2024-07-30

Abstract:Randomized rounding is a technique that was originally used to approximate hard offline discrete optimization problems from a mathematical programming relaxation. Since then it has also been used to approximately solve sequential stochastic optimization problems, overcoming the curse of dimensionality. To elaborate, one first writes a tractable linear programming relaxation that prescribes probabilities with which actions should be taken. Rounding then designs a (randomized) online policy that approximately preserves all of these probabilities, with the challenge being that the online policy faces hard constraints, whereas the prescribed probabilities only have to satisfy these constraints in expectation. Moreover, unlike classical randomized rounding for offline problems, the online policy's actions unfold sequentially over time, interspersed by uncontrollable stochastic realizations that affect the feasibility of future actions. This tutorial provides an introduction for using randomized rounding to design online policies, through four self-contained examples representing fundamental problems in the area: online contention resolution, stochastic probing, stochastic knapsack, and stochastic matching.

Data Structures and Algorithms

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper explores how to use the Randomized Rounding technique to design online algorithms in order to solve a series of online allocation, sorting, and matching problems. Specifically, the author attempts to solve the problems in the following ways:

1. **Overcoming the curse of dimensionality**: Traditional dynamic programming methods face the problem of exponential growth in computational complexity when dealing with high - dimensional state spaces, which is the "curse of dimensionality". The paper proposes a method based on Linear Programming Relaxation and Randomized Rounding, which can effectively approximate the solution of such problems.

2. **Dealing with uncertainty**: In the online decision - making process, the uncertainty of the future is inevitable. By introducing a probability model, the paper uses the Randomized Rounding technique to design an online strategy that can make approximately optimal decisions in an uncertain environment.

3. **Maintaining probability constraints**: In online decision - making, the probabilities of certain actions need to meet specific constraint conditions. For example, in the resource allocation problem, it must be ensured that the probability of each agent being accepted does not exceed its occurrence probability. The paper shows how to approximately maintain these probability constraints in the online strategy through the Randomized Rounding technique.

4. **Providing fairness guarantees**: In some application scenarios, such as food distribution, fairness is an important consideration. The method in the paper can not only maximize the expected return, but also ensure that the probability of each agent obtaining resources is not lower than a certain threshold, thereby achieving a certain degree of fairness.

### Specific problem examples

The paper uses four specific examples to illustrate the application of the Randomized Rounding technique:

1. **Online Contention Resolution Scheme (OCRS)**: It solves the problem of how to fairly allocate limited resources among multiple agents.
2. **Stochastic Probing and Search**: It studies how to make dynamic decisions in an uncertain environment, such as the order in which a recruiting company decides to send offers to candidates.
3. **Stochastic Knapsack**: It studies how to maximize the rewards for completing tasks in an uncertain environment, similar to the scheduling problem of completing multiple tasks within a limited time.
4. **Stochastic Matching**: It studies a variant of the maximum matching problem in graphs, which includes sequential decision - making and the occurrence of random events.

### Summary

In general, this paper aims to provide a new solution for online optimization problems through the Randomized Rounding technique, especially for those scenarios with high - dimensional state spaces and future uncertainties. Through this method, not only can the curse of dimensionality be effectively avoided, but also approximately optimal decisions can be made in an uncertain environment, and certain fairness and probability constraints can be guaranteed.