Learning to Accelerate Heuristic Searching for Large-Scale Maximum Weighted b-Matching Problems in Online Advertising
IJCAI, pp. 3437-3443, 2020.
EI
Keywords:
combinatorial optimizationbipartite b matchingmultichannel graph neuraleconomic marketad allocationMore(15+)
Weibo:
Abstract:
Bipartite b-matching is fundamental in algorithm design, and has been widely applied into economic markets, labor markets, etc. These practical problems usually exhibit two distinct features: large-scale and dynamic, which requires the matching algorithm to be repeatedly executed at regular intervals. However, existing exact and approxi...More
Code:
Data:
Introduction
- Bipartite b-matching is one of the fundamental problems in computer science and operations research.
- Typical applications include resource allocation problems, such as job/server allocation in cloud computing and product recommendation[De Francisci Morales et al, 2011] and advertisement allocation [Agrawal et al, 2018] in economic markets.
- It has been utilized as an algorithmic tool in a variety of domains, including document clustering [Dhillon, 2001], computer vision [Zanfir and Sminchisescu, 2018], and as a subroutine in machine learning algorithms.
- The goal of the ad allocation is to search for a maximum weighted b-matching: selecting a subset of edges with the maximum total scores while satisfying the cardinality constraints
Highlights
- Bipartite b-matching is one of the fundamental problems in computer science and operations research
- A bipartite graph connects a large set of consumers and a large set of ads
- We investigate whether we can leverage the representation capability of neural networks to transfer the knowledge learned from previous solved instances to accelerate the solution computing on similar new instances
- We propose a parallelizable and scalable learning based framework NeuSearcher to accelerate the solution computing for large-scale b-matching
- Our NeuSearcher with the designed multichannel graph neural network computes the same solutions at the fastest speed by reducing more than 50% computing time
- Our NeuSearcher transfers knowledge learned from previous solved instances to save more than 50% of the computing time
Methods
- GreedyMR is one of the fastest parallel algorithms in computing b-matching problems.
- The authors evaluate NeuSearcher on both open and industrial datasets.
- Due to the memory limit (128G), the authors cannot calculate the exact solution using Gurobi optimizer for the first 7 datasets.
- The authors compare the matching quality of the approximate algorithms relative to the exact solution on the other 3 open datasets (Amazon review data [He and McAuley, 2016] and MovieLens data [Harper and Konstan, 2016])
Results
- For problems with larger sizes, the Gurobi fails to compute an optimal solution due to the memory limit (128G).
- This indicates that faster approximate approaches are good alternatives in solving large-scale b-matching problems and the NeuSearcher achieves the state-of-the-art solution quality.
- The authors' NeuSearcher transfers knowledge learned from previous solved instances to save more than 50% of the computing time
Conclusion
- To the best of the knowledge, the authors are the first to integrate deep learning methods to accelerate solving practical large-scale b-matching problems.
- The authors' NeuSearcher transfers knowledge learned from previous solved instances to save more than 50% of the computing time.
- The authors design a parallel heuristic search algorithm to ensure the solution quality exactly the same with the state-of-the-art approximation algorithms.
- Experiments on open and real-world large-scale datasets show NeuSearcher can compute nearly optimal solution much faster than state-of-the-art methods
Summary
Introduction:
Bipartite b-matching is one of the fundamental problems in computer science and operations research.- Typical applications include resource allocation problems, such as job/server allocation in cloud computing and product recommendation[De Francisci Morales et al, 2011] and advertisement allocation [Agrawal et al, 2018] in economic markets.
- It has been utilized as an algorithmic tool in a variety of domains, including document clustering [Dhillon, 2001], computer vision [Zanfir and Sminchisescu, 2018], and as a subroutine in machine learning algorithms.
- The goal of the ad allocation is to search for a maximum weighted b-matching: selecting a subset of edges with the maximum total scores while satisfying the cardinality constraints
Methods:
GreedyMR is one of the fastest parallel algorithms in computing b-matching problems.- The authors evaluate NeuSearcher on both open and industrial datasets.
- Due to the memory limit (128G), the authors cannot calculate the exact solution using Gurobi optimizer for the first 7 datasets.
- The authors compare the matching quality of the approximate algorithms relative to the exact solution on the other 3 open datasets (Amazon review data [He and McAuley, 2016] and MovieLens data [Harper and Konstan, 2016])
Results:
For problems with larger sizes, the Gurobi fails to compute an optimal solution due to the memory limit (128G).- This indicates that faster approximate approaches are good alternatives in solving large-scale b-matching problems and the NeuSearcher achieves the state-of-the-art solution quality.
- The authors' NeuSearcher transfers knowledge learned from previous solved instances to save more than 50% of the computing time
Conclusion:
To the best of the knowledge, the authors are the first to integrate deep learning methods to accelerate solving practical large-scale b-matching problems.- The authors' NeuSearcher transfers knowledge learned from previous solved instances to save more than 50% of the computing time.
- The authors design a parallel heuristic search algorithm to ensure the solution quality exactly the same with the state-of-the-art approximation algorithms.
- Experiments on open and real-world large-scale datasets show NeuSearcher can compute nearly optimal solution much faster than state-of-the-art methods
Tables
- Table1: The structural properties of the datasets
- Table2: The solution quality comparison (best in bold)
- Table3: The runtimes (in seconds) of b-matching computation, where lower values are better (best in bold)
Funding
- The work is supported by the Alibaba Group through Alibaba Innovative Research Program, the National Natural Science Foundation of China (Grant Nos.: 61702362, U1836214) and the new Generation of Artificial Intelligence Science and Technology Major Project of Tianjin under grant: 19ZXZNGX00010
Reference
- [Agrawal et al., 2018] Shipra Agrawal, Morteza Zadimoghaddam, and Vahab Mirrokni. Proportional allocation: Simple, distributed, and diverse matching with high entropy. In International Conference on Machine Learning, pages 99–108, 2018.
- [Avis, 1983] David Avis. A survey of heuristics for the weighted matching problem. Networks, 13(4):475–493, 1983.
- [Bayati et al., 2011] Mohsen Bayati, Christian Borgs, Jennifer Chayes, and Riccardo Zecchina. Belief propagation for weighted b-matchings on arbitrary graphs and its relation to linear programs with integer solutions. SIAM Journal on Discrete Mathematics, 25(2):989–1011, 2011.
- [Chen and Tian, 2019] Xinyun Chen and Yuandong Tian. Learning to perform local rewriting for combinatorial optimization. In Advances in Neural Information Processing Systems, pages 6278–6289, 2019.
- [De Francisci Morales et al., 2011] Gianmarco De Francisci Morales, Aristides Gionis, and Mauro Sozio. Social content matching in mapreduce. Proceedings of the VLDB Endowment, 4(7):460–469, 2011.
- [Dhillon, 2001] Inderjit S Dhillon. Co-clustering documents and words using bipartite spectral graph partitioning. In Proceedings of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining, pages 269–274. ACM, 2001.
- [Ding et al., 2019] Jian-Ya Ding, Chao Zhang, Lei Shen, Shengyin Li, Bing Wang, Yinghui Xu, and Le Song. Optimal solution predictions for mixed integer programs. arXiv preprint arXiv:1906.09575, 2019.
- [Edmonds, 1965] Jack Edmonds. Maximum matching and a polyhedron with 0, 1-vertices. Journal of research of the National Bureau of Standards B, 69(125-130):55–56, 1965.
- [Grotschel and Holland, 1985] Martin Grotschel and Olaf Holland. Solving matching problems with linear programming. Mathematical Programming, 33(3):243–259, 1985.
- [Gurobi, 2014] Gurobi. Inc. gurobi optimizer reference manual, 2015. URL: http://www.gurobi.com, 2014.
- [Harper and Konstan, 2016] F Maxwell Harper and Joseph A Konstan. The movielens datasets: History and context. Acm transactions on interactive intelligent systems (tiis), 5(4):19, 2016.
- [He and McAuley, 2016] Ruining He and Julian McAuley. Ups and downs: Modeling the visual evolution of fashion trends with one-class collaborative filtering. In proceedings of the 25th international conference on world wide web, pages 507–517. International World Wide Web Conferences Steering Committee, 2016.
- [He et al., 2014] He He, Hal Daume III, and Jason M Eisner. Learning to search in branch and bound algorithms. In Advances in neural information processing systems, pages 3293–3301, 2014.
- [Hougardy, 2009] Stefan Hougardy. Linear time approximation algorithms for degree constrained subgraph problems. In Research Trends in Combinatorial Optimization, pages 185–200.
- [Khalil et al., 2017] Elias Khalil, Hanjun Dai, Yuyu Zhang, Bistra Dilkina, and Le Song. Learning combinatorial optimization algorithms over graphs. In Advances in Neural Information Processing Systems, pages 6348–6358, 2017.
- [Khan et al., 2016] Arif Khan, Alex Pothen, Md Mostofa Ali Patwary, Nadathur Rajagopalan Satish, Narayanan Sundaram, Fredrik Manne, Mahantesh Halappanavar, and Pradeep Dubey. Efficient approximation algorithms for weighted b-matching. SIAM Journal on Scientific Computing, 38(5):S593–S619, 2016.
- [Li et al., 2018] Zhuwen Li, Qifeng Chen, and Vladlen Koltun. Combinatorial optimization with graph convolutional networks and guided tree search. In Advances in Neural Information Processing Systems, pages 539–548, 2018.
- [Muller-Hannemann and Schwartz, 2000] Matthias MullerHannemann and Alexander Schwartz. Implementing weighted b-matching algorithms: insights from a computational study. Journal of Experimental Algorithmics (JEA), 5:8, 2000.
- [Naim and Manne, 2018] Md Naim and Fredrik Manne. Scalable b-matching on gpus. In 2018 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), pages 637–646. IEEE, 2018.
- [Padberg and Rao, 1982] Manfred W Padberg and M Ram Rao. Odd minimum cut-sets and b-matchings. Mathematics of Operations Research, 7(1):67–80, 1982.
- [Preis, 1999] Robert Preis. Linear time 1/2-approximation algorithm for maximum weighted matching in general graphs. In Annual Symposium on Theoretical Aspects of Computer Science, pages 259–269.
- [Vinyals et al., 2015] Oriol Vinyals, Meire Fortunato, and Navdeep Jaitly. Pointer networks. In Advances in Neural Information Processing Systems, pages 2692–2700, 2015.
- [Wu et al., 2019] Zonghan Wu, Shirui Pan, Fengwen Chen, Guodong Long, Chengqi Zhang, and Philip S Yu. A comprehensive survey on graph neural networks. arXiv preprint arXiv:1901.00596, 2019.
- [Xu et al., 2018] Keyulu Xu, Weihua Hu, Jure Leskovec, and Stefanie Jegelka. How powerful are graph neural networks? arXiv preprint arXiv:1810.00826, 2018.
- [Zanfir and Sminchisescu, 2018] Andrei Zanfir and Cristian Sminchisescu. Deep learning of graph matching. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2684–2693, 2018.
Tags
Comments