What are the interesting implementations of the auction algorithm?

There are also totally asynchronous implementations of the auction algorithm, which are interesting because they are quite flexible and may also result in faster solution.

What is the synchronization overhead of a Gauss-Seidel it?

This is significant because even though each Gauss-Seidel iteration may take less time because it is executed by multiple processors in parallel, the synchronization overhead is roughly proportional to the number of iterations.

What is the price of an object of a similarity class?

In particular, the price of an object j is the minimum contention threshold over the objects of the similarity class M(j) of jpj = min k∈M(j) p̂k. (24)Thus, all objects in a similarity class have the same price but possibly different contention thresholds.

Why can't the object prices decrease in the course of the auction algorithm?

The reason is that the object prices cannot decrease in the course of the algorithm, implying that yi ≥ aij − pj of all j other than j1 and j2.

What is the way to scale down the benefits of the artificial pairs?

Thus for problems which are normally expected to be feasible, it may be better to introduce artificial pairs with benefits that are of the order of −C, and then gradually scale downward these benefits towards the −(2n−1)C threshold if artificial pairs persist in the assignments obtained by the auction algorithm.

What is the length of the portion of P between i and k?

To see this, note that by Eq. (22b), pi − pk is the length of the portion of P between i and k, and every path connecting i to k must have length at least equal to pi − pk [add Eq. (22a) along the arcs of the path].

Open AccessJournal Article10.1007/BF00247653

Auction algorithms for network flow problems : a tutorial introduction

Dimitri P. Bertsekas

- 01 Oct 1992

- Computational Optimization and Applicati...

- Vol. 1, Iss: 1, pp 7-66

505

TL;DR: A new and comprehensive class of algorithms for solving the classical linear network flow problem and its various special cases such as shortest path, max-flow, assignment, transportation, and transhipment problems are surveyed.

Abstract: This paper surveys a new and comprehensive class of algorithms for solving the classical linear network flow problem and its various special cases such as shortest path, max-flow, assignment, transportation, and transhipment problems. The prototype method, from which the other algorithms can be derived, is the auction algorithm for the assignment problem. This is an intuitive method that operates like a real auction where persons compete for objects by raising their prices through competitive bidding; the prices can be viewed as dual variables. Conceptually, auction algorithms represent a significant departure from the cost improvement idea that underlies primal simplex and dual ascent methods; at any one iteration, they may deteriorate both the primal and the dual cost. Auction algorithms perform very well for several important types of problems, both in theory and in practice, and they are also well suited for parallel computation.

AI Agents for this Paper

Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps

Most frequently asked questions

1. What are the contributions in this paper?

This paper surveys a new and comprehensive class of algorithms for solving the classical linear network flow problem and its various special cases such as shortest path, max-flow, assignment, transportation, and transhipment problems.

2. What is the common way to initialize the algorithm so that the!-CS conditions are?

A common way to initialize the algorithm so that the !-CS conditions are satisfied is to take S to be empty, to choose p arbitrarily, and to select π as a function of p via the relation πi = maxk∈

3. What is the approach for constructing auction algorithms for short path problems?

Their approach for constructing auction algorithms for such problems is to convert them to assignment problems, and then to suitably apply the auction algorithm and streamline the computations.

4. What is the derivation of the auction algorithm for the assignment problem?

The derivation is based on well-known transformations of the general linear minimum cost flow problem and its various special cases (e.g., shortest path, max-flow, etc.) to equivalent assignment problems.

Figure 3: A sequence of prices p1 and p2 generated by the auction algorithm for the example of Figs. 1 and 2. The figure shows the equal dual cost surfaces in the space of p1 and p2, with p3 fixed at 0.

Figure 8: Typical example where a transportation problem is converted into an equivalent assignment problem and the duplicate objects induce price wars in the auction algorithm. Here a11, a21, and a31 are much larger than zero. The persons 1, 2, and 3 will keep on bidding up the prices of objects 1 and 1′ by 2" increments until one of the prices p1 or p1′ reaches or exceeds the minimum of a11, a21, and a31. (For a11 = a21 = a31 = C, this example is the same as the one of Figs. 1-3.)

Figure 7: Illustration of the conversion of a transportation problem into an assignment problem. Each source i is transformed into αi persons (for example source 1 is transformed into persons 1 and 1′), and each sink j is transformed into βj objects (for example sink 1 is transformed into objects 1, 1′, and 1′′). If (i, j) is an arc of benefit aij in the transportation problem, there is an assignment arc of benefit aij connecting each of the αi persons corresponding to i with each of the βj objects corresponding to j.

Figure 5: Illustration of the CS conditions for the shortest path problem. If each node is a ball, and for every arc (i, j) ∈ A, nodes i and j are connected with a string of length aij , the vertical coordinates pi of the nodes satisfy pi − pj ≤ aij , as shown in (b) for the problem given in (a). If the model is picked up and left to hang from the origin node 1, then p1 − pi gives the shortest distance to each node i, as shown in (c).

Figure 9: Illustration of the bid of person i in the auction algorithm with similar objects. The object j∗ offers the best value aij − p̂j for person i, based on the contention thresholds p̂j . However, here wi is the value offered by the second best similarity class, rather than the value of the second best object as in the regular auction algorithm. When the second best object belongs to the similarity class of j∗ (as in this figure), the bid of person i will be higher in the auction algorithm with similar objects that in the regular auction algorithm.

Figure 10: Illustration of the auction algorithm with similar persons. In this algorithm, all the persons in a similarity class with k persons submit bids for the k best objects for the class. The levels of the bids are such that once the bids are accepted, the profit of each person in the class is set at " below the profit offered by the (k+1)st best object. In the example of the figure the similar persons result from the conversion of the transportation source 1 into the two similar persons 1 and 1′. Consider first the case where the initial assignment is empty, all initial prices are 0, and a11 = a12 = a21 = a12 = C > 0. Then the regular auction algorithm will operate as in Fig. 2, involving a long price war if " is small relative to C. In the first iteration of the auction algorithm with similar persons, persons 1 and 1′ will submit a common bid of C + " for the two best objects 1 and 2. Then person 2 will submit a bid for object 3 at the next iteration and the algorithm will terminate, thus avoiding the price war. If instead we have a11 > a12 > 0, at the initial iteration persons 1 and 1′ will submit a bid of a11 + " for object 1 and a bid of a12 + " for object 2, bringing the net value of these objects to the value of the third object minus ". If a12 − " > a22 > a21 > 0, person 2 will then submit a bid for object 3 at the next iteration and the algorithm will terminate. If a22 > a12 − " and a22 > a21 > 0, person 3 will submit a bid for object 2 at the second iteration, raising its price to a22 + "; persons 1 and 1′ will then submit another common bid for objects 1 and 3 at the third iteration, and the algorithm will terminate.

Citations

•Book

Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations

Yoav Shoham, +1 more

- 15 Dec 2008

TL;DR: This exciting and pioneering new overview of multiagent systems, which are online systems composed of multiple interacting intelligent agents, i.e., online trading, offers a newly seen computer science perspective on multi agent systems, while integrating ideas from operations research, game theory, economics, logic, and even philosophy and linguistics.

...read moreread less

2.3K

•Posted Content

Computational Optimal Transport

Gabriel Peyré, +1 more

- 01 Mar 2018

- arXiv: Machine Learning

TL;DR: This short book reviews OT with a bias toward numerical methods and their applications in data sciences, and sheds lights on the theoretical properties of OT that make it particularly useful for some of these applications.

...read moreread less

1.9K

•Book

Network Optimization: Continuous and Discrete Models

Dimitri P. Bertsekas

- 01 May 1998

Journal Article•10.1016/J.TRB.2015.03.002

Planning, operation, and control of bus transport systems: A literature review

Omar J. Ibarra-Rojas, +3 more

- 01 Jul 2015

- Transportation Research Part B-methodolo...

TL;DR: In this paper, the authors present a comprehensive review of the literature on transit network planning problems and real-time control strategies suitable for bus transport systems, emphasizing recent studies as well as works not addressed in previous reviews.

...read moreread less

568

Journal Article•10.1006/GAME.2000.0822

Auction Protocols for Decentralized Scheduling

Michael P. Wellman, +3 more

- 01 Apr 2001

- Games and Economic Behavior

TL;DR: This work formalizes decentralized scheduling as a discrete resource allocation problem, and brings to bear some relevant economic concepts about the existence of equilibrium prices for some general classes of scheduling problems, and the quality of equilibrium solutions.

...read moreread less

537

...

Expand

References

•Journal Article•10.1002/NAV.3800020109

The Hungarian method for the assignment problem

Harold W. Kuhn

- 01 Mar 1955

- Naval Research Logistics Quarterly

TL;DR: This paper has always been one of my favorite children, combining as it does elements of the duality of linear programming and combinatorial tools from graph theory, and it may be of some interest to tell the story of its origin this article.

...read moreread less

14.3K

Journal Article•10.2307/2322374

Combinatorial optimization: algorithms and complexity

Christos H. Papadimitriou, +1 more

- 01 Jun 1982

- American Mathematical Monthly

TL;DR: This clearly written, mathematically rigorous text includes a novel algorithmic exposition of the simplex method and also discusses the Soviet ellipsoid algorithm for linear programming; efficient algorithms for network flow, matching, spanning trees, and matroids; the theory of NP-complete problems; approximation algorithms, local search heuristics for NPcomplete problems, more.

...read moreread less

7.6K

•Book

Parallel and Distributed Computation: Numerical Methods

Dimitri P. Bertsekas, +1 more

- 01 Jan 1989

TL;DR: This work discusses parallel and distributed architectures, complexity measures, and communication and synchronization issues, and it presents both Jacobi and Gauss-Seidel iterations, which serve as algorithms of reference for many of the computational approaches addressed later.

...read moreread less

•Book

Linear Programming and Extensions

George B. Dantzig

- 01 Jan 1963

TL;DR: This classic book looks at a wealth of examples and develops linear programming methods for their solutions and begins by introducing the basic theory of linear inequalities and describes the powerful simplex method used to solve them.

...read moreread less

6.1K

Journal Article•10.1109/TASSP.1984.1164450

Combinatorial optimization:Algorithms and complexity

Lawrence R. Rabiner

- 01 Dec 1984

- IEEE Transactions on Acoustics, Speech, ...

TL;DR: Eventually, you will unquestionably discover a supplementary experience and achievement by Spending more cash by spending more cash.

...read moreread less

...

Expand

Auction algorithms for network flow problems : a tutorial introduction

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What are the contributions in this paper?

2. What is the common way to initialize the algorithm so that the!-CS conditions are?

3. What is the approach for constructing auction algorithms for short path problems?

4. What is the derivation of the auction algorithm for the assignment problem?

5. What are the interesting implementations of the auction algorithm?

6. What is the synchronization overhead of a Gauss-Seidel it?

7. What is the price of an object of a similarity class?

8. Why can't the object prices decrease in the course of the auction algorithm?

9. What is the way to scale down the benefits of the artificial pairs?

10. What is the length of the portion of P between i and k?

Figures

Citations

Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations

Computational Optimal Transport

Network Optimization: Continuous and Discrete Models

Planning, operation, and control of bus transport systems: A literature review

Auction Protocols for Decentralized Scheduling

References

The Hungarian method for the assignment problem

Combinatorial optimization: algorithms and complexity

Parallel and Distributed Computation: Numerical Methods

Linear Programming and Extensions

Combinatorial optimization:Algorithms and complexity

Related Papers (5)

The Hungarian method for the assignment problem

Parallel and Distributed Computation: Numerical Methods

Algorithms for the Assignment and Transportation Problems

Counterspeculation, auctions, and competitive sealed tenders

Polar Factorization and Monotone Rearrangement of Vector-Valued Functions