Book cover

International Symposium on Algorithms and Computation

ISAAC 1992: Algorithms and Computation pp 351–369 Cite as

Generalized assignment problems

Session 7: Invited Papers

  • Silvano Martello 1 &
  • Paolo Toth 2  
  • Conference paper
  • First Online: 01 January 2005

410 Accesses

7 Citations

Part of the Lecture Notes in Computer Science book series (LNCS,volume 650)

We consider generalized assignment problems with different objective functions: min-sum, max-sum, min-max, max-min. We review transformations, bounds, approximation algorithms and exact algorithms. The results of extensive computational experiments are given.

  • Approximation Algorithm
  • Knapsack Problem
  • Resource Requirement
  • Lagrangian Relaxation
  • Decision Node

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

This is a preview of subscription content, log in via an institution .

Unable to display preview.  Download preview PDF.

G. Carpaneto, S. Martello, P. Toth (1988). Algorithms and codes for the assignment problem. In B. Simeone, P. Toth, G. Gallo, F. Maffioli, S. Pallottino (eds.). Fortran Codes For Network Optimization, Annals of Operations Research 13, 193–223.

Google Scholar  

A. De Maio, C. Roveda (1971). An all zero-one algorithm for a certain class of transportation problems. Operations Research 19, 1406–1418.

M.L. Fisher (1981). The Lagrangian relaxation method for solving integer programming problems. Management Science 27, 1–18.

M.L. Fisher, R. Jaikumar, L.N. VanWassenhove (1986). A multiplier adjustment method for the generalized assignment problem. Management Science 32, 1095–1103.

E.S. Gottlieb, M.R. Rao (1990a). The generalized assignment problem: valid inequalities and facets. Mathematical Programming 46, 31–52.

Article   Google Scholar  

E.S. Gottlieb, M.R. Rao (1990b). (1,k)-configuration facets for the generalized assignment problem. Mathematical Programming 46, 53–60.

M.M. Guignard, S. Kim (1987). Lagrangean decomposition: A model yielding stronger Lagrangean bounds. Mathematical Programming 39, 215–228.

M.M. Guignard, M.B. Rosenwein (1989). An improved dual based algorithm for the generalized assignment problem. Operations Research 37, 658–663.

T.D. Klastorin (1979). An effective subgradient algorithm for the generalized assignment problem. Computers and Operations Research 6, 155–164.

N.W. Kuhn (1955). The hungarian method for the assignment problem. Naval Research Logistics Quarterly 2, 83–97.

E.L. Lawler (1976). Combinatorial Optimization: Networks and Matroids , Holt, Rinehart and Winston, New York.

K. Jörnsten, M. NÄsberg (1986). A new Lagrangian relaxation approach to the generalized assignment problem. European Journal of Operational Research 27, 313–323.

S. Martello, P. Toth (1980). Solution of the zero-one multiple knapsack problem. European Journal of Operational Research 4, 276–283.

S. Martello, P. Toth (1981). An algorithm for the generalized assignment problem. In J.P. Brans (ed.), Operational Research'81 , North-Holland, Amsterdam, 589–603.

S. Martello, P. Toth (1990). Knapsack Problems: Algorithms and Computer Implementations , Wiley, Chichester.

S. Martello, P. Toth (1991).The bottleneck generalized assignment problem. Research report DEIS OR/5/91, University of Bologna.

J.B. Mazzola (1989). Generalized assignment with nonlinear capacity interaction. Management Science 35, 923–941.

J.B. Mazzola, A.W. Neebe (1988). Bottleneck generalized assignment problems. Engineering Costs and Production Economics 14, 61–65.

G.T. Ross, R.M. Soland (1975). A branch and bound algorithm for the generalized assignment problem. Mathematical Programming 8, 91–103.

V. Srinivasan, G.L. Thompson (1973). An algorithm for assigning uses to sources in a special class of transportation problems. Operations Research 21, 284–295.

Download references

Author information

Authors and affiliations.

Dipartimento di Informatica, Univcrsità di Torino, Italy

Silvano Martello

DEIS, Università di Bologna, Italy

You can also search for this author in PubMed   Google Scholar

Editor information

Rights and permissions.

Reprints and permissions

Copyright information

© 1992 Springer-Verlag Berlin Heidelberg

About this paper

Cite this paper.

Martello, S., Toth, P. (1992). Generalized assignment problems. In: Ibaraki, T., Inagaki, Y., Iwama, K., Nishizeki, T., Yamashita, M. (eds) Algorithms and Computation. ISAAC 1992. Lecture Notes in Computer Science, vol 650. Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-56279-6_88

Download citation

DOI : https://doi.org/10.1007/3-540-56279-6_88

Published : 09 June 2005

Publisher Name : Springer, Berlin, Heidelberg

Print ISBN : 978-3-540-56279-5

Online ISBN : 978-3-540-47501-9

eBook Packages : Springer Book Archive

Share this paper

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Publish with us

Policies and ethics

  • Find a journal
  • Track your research

Solving Generalized Assignment Problem using Branch-And-Price

Oct 27, 2021

Table of Contents

Generalized assignment problem, dantzig-wolfe decomposition, column generation, standalone model, initial solution, branching rule, branch-and-price.

One of the best known and widely researched problems in combinatorial optimization is Knapsack Problem:

  • given a set of items, each with its weight and value,
  • select a set of items maximizing total value
  • that can fit into a knapsack while respecting its weight limit.

The most common variant of a problem, called 0-1 Knapsack Problem can be formulated as follows:

  • \(m\) - number of items;
  • \(x_i\) - binary variable indicating whether item is selected;
  • \(v_i\) - value of each items;
  • \(w_i\) - weight of each items;
  • \(W\) - maximum weight capacity.

As often happens in mathematics, or science in general, an obvious question to ask is how the problem can be generalized. One of generalization is Generalized Assignment Problem. It answers question - how to find a maximum profit assignment of \(m\) tasks to \(n\) machines such that each task (\(i=0, \ldots, m\)) is assigned to exactly one machine (\(j=1, \ldots, n\)), and one machine can have multiple tasks assigned to subject to its capacity limitation. Its standard formulation is presented below:

  • \(n\) - number of machines;
  • \(m\) - number of tasks;
  • \(x_{ij}\) - binary variable indicating whether task \(i\) is assigned to machine \(j\);
  • \(v_{ij}\) - value/profit of assigning task \(i\) to machine \(j\);
  • \(w_{ij}\) - weight of assigning task \(i\) to machine \(j\);
  • \(c_j\) - capacity of machine \(j\).

Branch-and-price

Branch-and-price is generalization of branch-and-bound method to solve integer programs (IPs),mixed integer programs (MIPs) or binary problems. Both branch-and-price, branch-and-bound, and also branch-and-cut, solve LP relaxation of a given IP. The goal of branch-and-price is to tighten LP relaxation by generating a subset of profitable columns associated with variables to join the current basis.

Branch-and-price builds at the top of branch-and-bound framework. It applies column generation priori to branching. Assuming maximization problem, branching occurs when:

  • Column Generation is finished (i.e. no profitable columns can be found).
  • Objective value of the current solution is greater than best lower bound.
  • The current solution does not satisfy integrality constraints.

However, if only first two conditions are met but not the third one, meaning the current solution satisfies integrality constraints, then the best solution and lower bound are updated (lower bound is tightened) with respectively the current solution and its objective value.

The crucial element needed to apply branch-and-price successfully is to find branching scheme. It is tailored to specific problem to make sure that it does not destroy problem structure and can be used in pricing subproblem to effectively generate columns that enter Restricted Master Problem (RMP) while respecting branching rules .

Below is flow diagram describing branch-and-price method:

Branch-and-Price flow diagram

The successful application B&P depends on tight/strong model formulation. Model formulation is considered tight if solution of its LP relaxation satisfies (frequently) integrality constraints. One of structured approaches to come up with such a formulation is to use Dantzig-Wolfe Decomposition technique. We will see example of it applied to Generalized Assignment Problem (GAP).

A standard formulation was described above. Now, let’s try to reformulate problem. Let

be a set containing all feasible solutions to Knapsack problem for \(j\)-th machine. Clearly, \(S_j\) contains finite number of points, so \(S_j = \{ \mathbf{z}_j^1, \ldots, \mathbf{z}_j^{K_j} \}\), where \(\mathbf{z}_j^k \in \{0, 1\}^{m}\). You can think about \(\mathbf{z}_j^k \in \{0, 1\}^{m}\) as 0-1 encoding of tasks that form \(k\)-th feasible solution for machine \(j\). Now, let \(S = \{ \mathbf{z}_1^1, \ldots, \mathbf{z}_1^{K_1}, \ldots, \mathbf{z}_n^1, \ldots, \mathbf{z}_n^{K_n} \}\) be a set of all feasible solution to GAP. It, potentially, contains a very large number of elements. Then, every point \(x_{ij}\) can be expressed by the following convex combination:

where \(z_{ij}^k \in \{0, 1\}\), and \(z_{ij}^k = 1\) iff task \(i\) is assigned to machine \(j\) in \(k\)-th feasible solution for the machine.

Now, let’s use this representation to reformulate GAP:

Note that we do not need capacity restrictions as they are embedded into definition of feasible solution for machine \(j\).

Now that we have formulation that is suitable for column generation, let’s turn our attention to it.

Column generation is another crucial component of branch-and-price. There are many great resources devoted to column generation so I will mention only core points:

  • Column generation is useful when a problem’s pool of feasible solutions contains many elements but only small subset will be present in the optimal solution.
  • There exists subproblem (called often pricing problem) that can be used to effectively generate columns that should enter RMP.
  • Column generation starts with initial feasible solution.
  • Pricing subproblem objective function is updated with dual of the current solution.
  • Columns with positive reduced cost, in case of maximization problem, enter problem.
  • Procedure continues until such columns exist.

Below is flow diagram describing column generation method:

Column generation flow diagram

Implementation

Let’s see how one can approach implementation of B&P to solve Generalized Assignment Problem. Below is discussion about main concepts and few code excerpts, a repository containing all code can be found on github .

An example of problem instance taken from [1] is:

It is always good idea to have a reference simple(r) implementation that can be used to validate our results using more sophisticated methods. In our case it is based on standard problem formulation. Implementation can be found in repo by checking classes GAPStandaloneModelBuilder and GAPStandaloneModel . Formulation for a problem instance presented above looks as follows:

Now let’s try to examine building blocks of B&P to discus main part at the end, once all the puzzles are there.

To start column generation process, we need to have an initial solution. One possible way to derive it is to use two-phase Simplex method. In first step, you add slack variables to each constraint and set objective function as their sum. Then you minimize the problem. If your solution has objective value \(0\), then first of all you have initial solution and you know that your problem is feasible. In case you end up with positive value for any of slack variables, you can conclude that the problem is infeasible. You can stop here.

I took a different approach and came up with simple heuristic that generate initial solution. I have not analyzed it thoroughly so I am not sure if it is guaranteed to always return feasible solution if one exists. Its idea is quite simple:

  • Construct bipartite graph defined as \(G=(V, A)\), where \(V = T \cup M\) – \(T\) is set of tasks and obviously \(M\) is set of machines. There exists arc \(a = (t, m)\) if \(w_{tm} \le rc_{m}\), where \(rc_{m}\) is remaining capacity for machine \(m\). Initially remaining capacity is equal to capacity of machine and with each iteration, and assignment of task to machine it is being update. If \(\vert A \vert = 0\), then stop.
  • Solve a minimum weight matching problem.
  • Update assignments – say that according to solution task \(t_0\) should be assigned to machine \(m_0\), then \(\overline{rc}_{m_0} = rc_{m_0} - w_{t_0 m_0}\).
  • Find a machine where task is contributing with the lowest weight – say machine \(m_0 = \arg\min \{ m: w_{t_0 m} \}\).
  • Free up remaining capacity so there is enough space for \(t_0\) on machine \(m_0\). Any tasks that were de-assigned in a process are added to pool of unassigned tasks.
  • Repeat until there are no unassigned tasks.

See details on github .

As we said before the most important piece needed to implement B&P is branching rules which does not destroy structure of subproblem. Let’s consider non-integral solution to RMP. Given convexity constraint it means that there exists machine \(j_0\) and at least two, and for sake of example say exactly two, \(0 < \lambda_{j_0} ^{k_1} < 1\) and \(0 < \lambda_{j_0} ^{k_2} < 1\) such that \(\lambda_{j_0} ^{k_1} + \lambda_{j_0} ^{k_2} = 1\). Since with each of \(\lambda\)s is connected different assignment (set of tasks), then it leads us to a conclusion that there exists task \(i_0\) such that \(x_{i_0 j_0} < 1\) expressed in variables from the original formulation. Now, let’s use this information to formulate branching rule:

  • left child node: a task \(i_0\) must be assigned to a machine \(j_0\).
  • right child node: a task \(i_0\) cannot be assigned to a machine \(j_0\).

We can say that branching is based on \(x_{ij}\) from standard formulation. And it can be represented by:

Note that we can use the branching rule to easily to filter out initial columns for each node that do not satisfy those conditions:

  • \(j = j_0\) and task \(i_0 \in T_{j_0}\), or
  • \(j \neq j_0\) and task \(i_0 \notin T_{j}\).
  • \(j = j_0\) and task \(i_0 \in T_{j_0}\).

See on github .

Based on the same principle, subproblem’s pool of feasible solution are created - i.e. on left child node:

  • knapsack subproblem for machine \(j_0\) – variable representing task \(i_0\) is forced to be \(1\).
  • knapsack subproblem for machine \(j \neq j_0\) – variable representing task \(i_0\) is forced to be \(0\).

Similarly for right childe node. See on github .

Below is an outline of main loop of column generation. It is an implementation of flow diagram from above so I will not spend too much time describing it. The only part maybe worth commenting is stop_due_to_no_progress - it evaluates whether column generation did not make any progress in last \(k\)-iterations and it should be stop.

Now, let’s see how constructing subproblems, solving them and then adding back column(s) to RMP looks like. We have as many subproblems as machines. Once a solution is available, we check whether it has positive reduced cost. A solution to knapsack problem corresponds to column in RMP. So if the column with positive reduced cost was identified and added, then new iteration of column generation will be executed. Gurobi allows to query information about all other identified solutions, so we can utilize this feature and add all columns that have the same objective value as optimal solution, potentially adding more than one column and hoping it will positively impact solution time.

Note that each subproblem is independent so in principle they could be solved in parallel. However due to Python Global Interpreter Lock (GIL) that prevent CPU-bounded threads to run in parallel, they are solved sequentially. Additionally depending on your Gurobi license, you might not be allowed to solve all those models in parallel even if Python would allow it.

Below you can find example of one of the RMPs:

and subproblem with dual information passed:

Now that we have all building blocks prepared, then let’s turn our attention back to B&P.

In the blog post, Branch-and-Price technique for solving MIP was explained. An example of applying B&P for Generalized Assignment Problem was presented. The solution approach used Python as programming language and Gurobi as solver.

[1] Der-San Chen, Robert G. Batson, Yu Dang (2010), Applied Integer Programming - Modeling and Solution, Willey. [2] Lasdon, Leon S. (2002), Optimization Theory for Large Systems, Mineola, New York: Dover Publications. [3] Cynthia Barnhart, Ellis L. Johnson, George L. Nemhauser, Martin W. P. Savelsbergh, Pamela H. Vance, (1998) Branch-and-Price: Column Generation for Solving Huge Integer Programs. Operations Research 46(3):316-329.

Help | Advanced Search

Computer Science > Computer Vision and Pattern Recognition

Title: good: towards domain generalized orientated object detection.

Abstract: Oriented object detection has been rapidly developed in the past few years, but most of these methods assume the training and testing images are under the same statistical distribution, which is far from reality. In this paper, we propose the task of domain generalized oriented object detection, which intends to explore the generalization of oriented object detectors on arbitrary unseen target domains. Learning domain generalized oriented object detectors is particularly challenging, as the cross-domain style variation not only negatively impacts the content representation, but also leads to unreliable orientation predictions. To address these challenges, we propose a generalized oriented object detector (GOOD). After style hallucination by the emerging contrastive language-image pre-training (CLIP), it consists of two key components, namely, rotation-aware content consistency learning (RAC) and style consistency learning (SEC). The proposed RAC allows the oriented object detector to learn stable orientation representation from style-diversified samples. The proposed SEC further stabilizes the generalization ability of content representation from different image styles. Extensive experiments on multiple cross-domain settings show the state-of-the-art performance of GOOD. Source code will be publicly available.

Submission history

Access paper:.

  • Download PDF
  • HTML (experimental)
  • Other Formats

References & Citations

  • Google Scholar
  • Semantic Scholar

BibTeX formatted citation

BibSonomy logo

Bibliographic and Citation Tools

Code, data and media associated with this article, recommenders and search tools.

  • Institution

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs .

IMAGES

  1. (PDF) Solving the Generalized Assignment Problem: An Optimizing and

    generalized assignment method

  2. (PDF) A Study of Team Recommended Generalized Assignment Methods

    generalized assignment method

  3. (PDF) An Ejection Chain Approach for the Generalized Assignment Problem

    generalized assignment method

  4. PPT

    generalized assignment method

  5. (PDF) A (1–)-approximation algorithm for the generalized assignment problem

    generalized assignment method

  6. (PDF) Bicriteria Multiresource Generalized Assignment Problem

    generalized assignment method

VIDEO

  1. Individual Assignment: Statistic method for accounting and finance

  2. A generalized design method for learning-based disturbance observer

  3. 24 Generalization & Specialization

  4. Assignment Problem Degeneracy part -2

  5. Generalization and Specialization in DBMS

  6. Lec-23 Unbalanced

COMMENTS

  1. Generalized assignment problem

    In applied mathematics, the maximum generalized assignment problem is a problem in combinatorial optimization. This problem is a generalization of the assignment problem in which both tasks and agents have a size. Moreover, the size of each task might vary from one agent to the other.

  2. Generalized Assignment Problem

    The generalized assignment problem (GAP) seeks the minimum cost assignment of n tasks to m agents such that each task is assigned to precisely one agent subject to capacity restrictions on the agents. The formulation of the problem is: $$ \min\hspace {.4in}\sum_ {i=1}^m \sum_ {j=1}^n c_ {ij}x_ {ij}\: $$ (1)

  3. PDF The Generalized Assignment Problem and Its Generalizations

    The generalized assignment problem is a classical combinatorial optimization problem that models a variety of real world applications including flexible manufacturing systems [6], facility location [11] and vehicle routing problems [2]. Given n jobs J = 1, 2, . . . , n and m agents { }

  4. (PDF) Solving the Generalized Assignment Problem: An ...

    The classical generalized assignment problem (GAP) may be stated as finding a minimum-cost assignment of tasks to agents such that each task is assigned to exactly one agent and such that...

  5. Generalized Assignment Problem

    In the Generalized Assignment Problem (GAP), one is given items and knapsacks of capacity : inserting item into knapsack produces a profit and assigns a weight to knapsack . The objective is to pack each item into exactly one knapsack so as to maximize the overall profit without assigning to any knapsack a total weight greater than its capacity.

  6. Solving the reliability-oriented generalized assignment problem by

    The well-known generalized assignment problem (GAP) is to achieve the optimal assignment between agents and tasks where each task is assigned to an agent exactly once and the capacity restrictions are respected.

  7. [PDF] Generalized Assignment Problems

    TLDR. A truthful-in-expectation, (1-1/e)-approximation mechanism for a strategic variant of the generalized assignment problem (GAP), which improves upon the existing optimization algorithms for GAP in terms of simplicity and runtime while the approximation ratio closely matches the best approximation ratio known for G AP when all inputs are ...

  8. An exact method with variable fixing for solving the generalized

    We propose a simple exact algorithm for solving the generalized assignment problem. Our contribution is twofold: we reformulate the optimization problem into a sequence of decision problems, and we apply variable-fixing rules to solve these effectively. The decision problems are solved by a simple depth-first lagrangian branch-and-bound method, improved by our variable-fixing rules to prune ...

  9. Generalized assignment problems

    We consider generalized assignment problems with different objective functions: min-sum, max-sum, min-max, max-min. We review transformations, bounds, approximation algorithms and exact algorithms. The results of extensive computational experiments are given. Keywords Approximation Algorithm Knapsack Problem Resource Requirement

  10. Solving the Generalized Assignment Problem: An Optimizing and Heuristic

    A special purpose branch-and-bound algorithm that utilizes linear programming cuts, feasible-solution generators, Lagrangean relaxation, and subgradient optimization that can be used quite effectively as a heuristic when proof of optimality is not an absolute requirement is described. The classical generalized assignment problem (GAP) may be stated as finding a minimum-cost assignment of tasks ...

  11. A Study of Team Recommended Generalized Assignment Methods

    Firstly, a formal description of the team recommendation problem is given; secondly, a team-recommended generalized assignment model (TRGAM) is established based on the work ability value of alternative members, the comprehensive work ability value of the team as the core concern index, the importance weight of team tasks and the energy allocati...

  12. Effective algorithm and heuristic for the generalized assignment

    A standard subgradient method (SM), used at each node of the decision tree to solve the Lagrangian dual, provides an upper bound. Our key contribution in this paper is a new heuristic, applied at each iteration of the SM, which tries to exploit the solution of the relaxed problem, by solving a smaller generalized assignment problem.

  13. PDF Study of Team Recommended Generalized Assignment Methods

    into the assignment problem; (b) the team recommended generalized assignment model (abbreviated as TRGAM) is proposed; (c) give a basic guideline for refining alternative members; (d) a solution method based on the enumeration method and the Hungarian algorithm (abbreviated as BEM HM-TRGAMs) is designed for the standard TRGAM; and

  14. Assignment problem

    The assignment problem consists of finding, in a weighted bipartite graph, a matching of a given size, in which the sum of weights of the edges is minimum. If the numbers of agents and tasks are equal, then the problem is called balanced assignment. Otherwise, it is called unbalanced assignment. [1]

  15. A new model of generalized assignment problem and its method

    A new model of generalized assignment problem and its method Abstract: The traditional assignment problem assumes that the number of persons is equal to that of the tasks, beyond that, one person is allowed to undertake only one task and one task must be accomplished by one person.

  16. [PDF] A branch and bound algorithm for the generalized assignment

    A branch and bound algorithm is developed that solves the generalized assignment problem by solving a series of binary knapsack problems to determine the bounds. This paper describes what is termed the "generalized assignment problem". It is a generalization of the ordinary assignment problem of linear programming in which multiple assignments of tasks to agents are limited by some ...

  17. A Multiplier Adjustment Method for the Generalized Assignment Problem

    We describe a branch and bound algorithm for the generalized assignment problem in which bounds are obtained from a Lagrangian relaxation with the multipliers set by a heuristic adjustment method. The algorithm was tested on a large sample of small random problems and a number of large problems derived from a vehicle routing application.

  18. (PDF) Heuristics for the generalised assignment problem: Simulated

    A review of exact and heuristic methods is presented. A λ-generation mechanism is introduced. ... We study to solve our generalized assignment problems in the Bregman Alternating Direction Method ...

  19. A survey of algorithms for the generalized assignment problem

    M.L. Fisher, R., L. A multiplier adjustment method for the generalized assignment problem. Management Science, 32/9 ( 1986) 1095 - 1103. CrossRefGoogle Scholar. Gavish and Pirkul, 1985a. B. Gavish, H. Pirkul. Zero-one integer programs with few constraints — efficient branch and bound algorigthms.

  20. Solving Generalized Assignment Problem using Branch-And-Price

    Branch-and-price is generalization of branch-and-bound method to solve integer programs (IPs),mixed integer programs (MIPs) or binary problems. ... An example of applying B&P for Generalized Assignment Problem was presented. The solution approach used Python as programming language and Gurobi as solver. References [1] Der-San Chen, Robert G ...

  21. A Multiplier Adjustment Method for the Generalized Assignment Problem

    <P>We describe a branch and bound algorithm for the generalized assignment problem in which bounds are obtained from a Lagrangian relaxation with the multipliers set by a heuristic adjustment method. The algorithm was tested on a large sample of small random problems and a number of large problems derived from a vehicle routing application.

  22. A new classification method using the generalized basic probability

    An improved method is proposed to model the incomplete information where the generalized basic probability assignment (GBPA) is generated by using the Gaussian distribution model. First, we constructed the Gaussian distribution based on the mean and variance calculated from the training set.

  23. A multiplier adjustment method for the generalized assignment problem

    A branch and bound algorithm for the generalized assignment problem in which bounds are obtained from a Lagrangian relaxation with the multiplier adjustment method appears to be about one order of magnitude faster than the best previously existing algorithms for this problem. We describe a branch and bound algorithm for the generalized assignment problem in which bounds are obtained from a ...

  24. [2402.13064] Synthetic Data (Almost) from Scratch: Generalized

    We introduce Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs). Unlike prior work that relies on seed examples or existing datasets to construct instruction tuning data, GLAN exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction ...

  25. GOOD: Towards Domain Generalized Orientated Object Detection

    Oriented object detection has been rapidly developed in the past few years, but most of these methods assume the training and testing images are under the same statistical distribution, which is far from reality. In this paper, we propose the task of domain generalized oriented object detection, which intends to explore the generalization of oriented object detectors on arbitrary unseen target ...