Problem-specific Techniques Research Articles

Deep Learning for Consumer Devices and Services 3—Getting More From Your Datasets With Data Augmentation

A key aspect to developing and successfully deploying neural network (NN)-based solutions is the availability of suitable datasets. In this article some of the challenges to acquire and annotate data are discussed in the context of new consumer devices. To increase the sample size of training data several approaches to augment a seed dataset are explained and discussed including a number of advanced, problem-specific techniques. A basic introduction to the concept of learned data augmentation is also provided.

Tumor growth modeling: Parameter estimation with Maximum Likelihood methods

Background & objective: In this work, we focus on estimating the parameters of the widely used Gompertz tumor growth model, based on measurements of the tumor’s volume. Being able to accurately describe the dynamics of tumor growth on an individual basis is very important both for growth prediction and designing personalized, optimal therapy schemes (e.g. when using model predictive control).Methods: Our analysis aims to compute both the growth rate and the carrying capacity of the Gompertz function, along with the characteristics of the additive Gaussian process and measurement noise of the system. Three methods based on Maximum Likelihood estimation are proposed. The first utilizes an assumption regarding the measurement noise that simplifies the problem, the second combines the Extended Kalman Filter and Maximum Likelihood estimation, and the third is a nonstandard exact form of Maximum Likelihood estimation, where numerical integration is used to approximate the likelihood of the measurements, along with a novel way to reduce the required computations.Results: Synthetic data were used in order to perform extensive simulations aiming to compare the methods’ effectiveness, with respect to the accuracy of the estimation. The proposed methods are able to estimate the growth dynamics, even when the noise characteristics are not estimated accurately. Another very important finding is that the methods perform best in the case that corresponds to the problem needed to be solved when dealing with experimental data.Conclusion: Using nonstandard, problem specific techniques can improve the estimation accuracy and best exploit the available data.

Vehicle routing with roaming delivery locations

We propose the vehicle routing problem with roaming delivery locations (VRPRDL) to model an innovation in last-mile delivery where a customer’s order is delivered to the trunk of his car. We develop construction and improvement heuristics for the VRPRDL based on two problem-specific techniques: (1) efficiently optimizing the delivery locations for a fixed customer delivery sequence and (2) efficiently switching a predecessor’s or successor’s delivery location during the insertion or deletion of a customer in a route. Furthermore, we conduct an extensive computation study to assess and quantify the benefits of trunk delivery in a variety of settings. The study reveals that a significant reduction in total distance travelled can be achieved, especially when trunk delivery is combined with traditional home delivery, which has both economic and environmental benefits.

Multi-objective evolutionary fuzzy clustering for image segmentation with MOEA/D

In order to achieve robust performance of preserving significant image details while removing noise for image segmentation, this paper presents a multi-objective evolutionary fuzzy clustering (MOEFC) algorithm to convert fuzzy clustering problems for image segmentation into multi-objective problems. The multi-objective problems are optimized by multi-objective evolutionary algorithm with decomposition. The decomposition strategy is adopted to project the multi-objective problem into a number of sub-problems. Each sub-problem represents a fuzzy clustering problem incorporating local information for image segmentation. Opposition-based learning is utilized to improve search capability of the proposed algorithm. Two problem-specific techniques, an adaptive weighted fuzzy factor and a mixed population initialization, are introduced to improve the performance of the algorithm. Experiment results on synthetic and real images illustrate that the proposed algorithm can achieve a trade-off between preserving image details and removing noise for image segmentation.

Modified orbital branching for structured symmetry with an application to unit commitment

The past decade has seen advances in general methods for symmetry breaking in mixed-integer linear programming. These methods are advantageous for general problems with general symmetry groups. Some important classes of mixed integer linear programming problems, such as bin packing and graph coloring, contain highly structured symmetry groups. This observation has motivated the development of problem-specific techniques. In this paper we show how to strengthen orbital branching in order to exploit special structures in a problem's symmetry group. The proposed technique, to which we refer as modified orbital branching, is able to solve problems with structured symmetry groups more efficiently. One class of problems for which this technique is effective is when the solution variables can be expressed as 0/1 matrices where the problem's symmetry group contains all permutations of the columns. We use the unit commitment problem, an important problem in power systems, to demonstrate the strength of modified orbital branching.

Short sales in Log-robust portfolio management

This paper extends the Log-robust portfolio management approach to the case with short sales, i.e., the case where the manager can sell shares he does not yet own. We model the continuously compounded rates of return, which have been established in the literature as the true drivers of uncertainty, as uncertain parameters belonging to polyhedral uncertainty sets, and maximize the worst-case portfolio wealth over that set in a one-period setting. The degree of the manager’s aversion to ambiguity is incorporated through a single, intuitive parameter, which determines the size of the uncertainty set. The presence of short-selling requires the development of problem-specific techniques, because the optimization problem is not convex. In the case where assets are independent, we show that the robust optimization problem can be solved exactly as a series of linear programming problems; as a result, the approach remains tractable for large numbers of assets. We also provide insights into the structure of the optimal solution. In the case of correlated assets, we develop and test a heuristic where correlation is maintained only between assets invested in. In computational experiments, the proposed approach exhibits superior performance to that of the traditional robust approach.

Partitioned time discretization for parallel solution of coupled ODE systems

One decoupling method for multiphysics, multiscale, multidomain applications involves partitioning the problem via explicit time discretizations in the coupling terms. Specialized, problem-specific techniques are needed for the resulting partitioned methods to avoid time step restrictions which make long time calculations costly. This report studies unconditionally stable, uncoupled time stepping methods for a model problem sharing mathematical structure akin to the coupled atmosphere-ocean system. Three decoupled time stepping algorithms are introduced and their stability and consistency are rigorously examined. Numerical experiments are performed that study their stability and convergence properties.

A general audio classifier based on human perception motivated model

The audio channel conveys rich clues for content-based multimedia indexing. Interesting audio analysis includes, besides widely known speech recognition and speaker identification problems, speech/music segmentation, speaker gender detection, special effect recognition such as gun shots or car pursuit, and so on. All these problems can be considered as an audio classification problem which needs to generate a label from low audio signal analysis. While most audio analysis techniques in the literature are problem specific, we propose in this paper a general framework for audio classification. The proposed technique uses a perceptually motivated model of the human perception of audio classes in the sense that it makes a judicious use of certain psychophysical results and relies on a neural network for classification. In order to assess the effectiveness of the proposed approach, large experiments on several audio classification problems have been carried out, including speech/music discrimination in Radio/TV programs, gender recognition on a subset of the switchboard database, highlights detection in sports videos, and musical genre recognition. The classification accuracies of the proposed technique are comparable to those obtained by problem specific techniques while offering the basis of a general approach for audio classification.

Using statistical analysis to tune an evolutionary algorithm for dynamic optimization with progressive step reduction

We describe the use of an evolutionary algorithm (EA) to solve dynamic control optimization problems in engineering. In this class of problems, a set of control variables must be manipulated over time to optimize the outcome, which is obtained by solving a set of differential equations for the state variables. A problem-specific technique, progressive step reduction (PSR), is shown to considerably improve the efficiency of the algorithm for this application. Factorial experimentation and rigorous statistical analysis are used to determine the effects of PSR and tune the parameters of the algorithm.

A Tabu-Search Hyperheuristic for Timetabling and Rostering

Hyperheuristics can be defined to be heuristics which choose between heuristics in order to solve a given optimisation problem. The main motivation behind the development of such approaches is the goal of developing automated scheduling methods which are not restricted to one problem. In this paper we report the investigation of a hyperheuristic approach and evaluate it on various instances of two distinct timetabling and rostering problems. In the framework of our hyperheuristic approach, heuristics compete using rules based on the principles of reinforcement learning. A tabu list of heuristics is also maintained which prevents certain heuristics from being chosen at certain times during the search. We demonstrate that this tabu-search hyperheuristic is an easily re-usable method which can produce solutions of at least acceptable quality across a variety of problems and instances. In effect the proposed method is capable of producing solutions that are competitive with those obtained using state-of-the-art problem-specific techniques for the problems studied here, but is fundamentally more general than those techniques.

A Technical Review of Column Generation in Integer Programming

This paper provides a technical review of topics relevant to applying column generation methods to solve integer programs but emphasizes formulation issues as a means of achieving its goal, which is to bridge the gap between methodological development and application. Type I, II and III column generation approaches are described in detail and each is demonstrated by a set of prototypical formulations that provide a historical perspective of milestone contributions. Technical issues, including formulation, context, algorithm design and implementation are also related. Formulation issues encompass the restricted master problem (RMP) and subproblem (SP) structure, symmetry, complexity and the Integrality Property. Context issues comprise theoretical principles, dealing with binary or general integer variables and generating rows as well as columns. Algorithm design issues include branching strategies, SP solution strategies and problem-specific techniques. Implementation issues include determining an initial basic feasible solution, managing a pool of generated columns, optimizing the RMP at each iteration, and handling degeneracy and tailing off.