Computer Sciences Commons^™

Mnemonics Training: Multi-Class Incremental Learning Without Forgetting, Yaoyao Liu, Yuting Su, An-An Liu, Bernt Schiele, Qianru Sun

Research Collection School Of Computing and Information Systems

Multi-Class Incremental Learning (MCIL) aims to learn new concepts by incrementally updating a model trained on previous concepts. However, there is an inherent trade-off to effectively learning new concepts without catastrophic forgetting of previous ones. To alleviate this issue, it has been proposed to keep around a few examples of the previous concepts but the effectiveness of this approach heavily depends on the representativeness of these examples. This paper proposes a novel and automatic framework we call mnemonics, where we parameterize exemplars and make them optimizable in an end-to-end manner. We train the framework through bilevel optimizations, i.e., model-level and …

Revised Polyhedral Conic Functions Algorithm For Supervised Classification, Gürhan Ceylan, Gürkan Öztürk

Turkish Journal of Electrical Engineering and Computer Sciences

In supervised classification, obtaining nonlinear separating functions from an algorithm is crucial for prediction accuracy. This paper analyzes the polyhedral conic functions (PCF) algorithm that generates nonlinear separating functions by only solving simple subproblems. Then, a revised version of the algorithm is developed that achieves better generalization and fast training while maintaining the simplicity and high prediction accuracy of the original PCF algorithm. This is accomplished by making the following modifications to the subproblem: extension of the objective function with a regularization term, relaxation of a hard constraint set and introduction of a new error term. Experimental results show that …

Computational Model For Neural Architecture Search, Ram Deepak Gottapu

Doctoral Dissertations

"A long-standing goal in Deep Learning (DL) research is to design efficient architectures for a given dataset that are both accurate and computationally inexpensive. At present, designing deep learning architectures for a real-world application requires both human expertise and considerable effort as they are either handcrafted by careful experimentation or modified from a handful of existing models. This method is inefficient as the process of architecture design is highly time-consuming and computationally expensive.

The research presents an approach to automate the process of deep learning architecture design through a modeling procedure. In particular, it first introduces a framework that treats …

Catgame: A Tool For Problem Solving In Complex Dynamic Systems Using Game Theoretic Knowledge Distribution In Cultural Algorithms, And Its Application (Catneuro) To The Deep Learning Of Game Controller, Faisal Waris

Wayne State University Dissertations

Cultural Algorithms (CA) are knowledge-intensive, population-based stochastic optimization methods that are modeled after human cultures and are suited to solving problems in complex environments. The CA Belief Space stores knowledge harvested from prior generations and re-distributes it to future generations via a knowledge distribution (KD) mechanism. Each of the population individuals is then guided through the search space via the associated knowledge. Previously, CA implementations have used only competitive KD mechanisms that have performed well for problems embedded in static environments. Relatively recently, CA research has evolved to encompass dynamic problem environments. Given increasing environmental complexity, a natural question arises …

Optimizing Pollution Routing Problem, Shivika Dewan

All Master's Theses

Pollution is a major environmental issue around the world. Despite the growing use and impact of commercial vehicles, recent research has been conducted with minimizing pollution as the primary objective to be reduced. The objective of this project is to implement different optimization algorithms to solve this problem. A basic model is created using the Vehicle Routing Problem (VRP) which is further extended to the Pollution Routing Problem (PRP). The basic model is updated using a Monte Carlo Algorithm (MCA). The data set contains 180 data files with a combination of 10, 15, 20, 25, 50, 75, 100, 150, and …

Optimizing Cluster Sets For The Scan Statistic Using Local Search, James Shulgan

Graduate Research Theses & Dissertations

In recent years, scattering sensors to produce wireless sensor networks (WSN) has been proposed for detecting localized events in large areas. Because sensor measurements are noisy, the WSN needs to use statistical methods such as the scan statistic. The scan statistic groups measurements into various clusters, computes a cluster statistic for each cluster, and decides that an event has happened if any of the statistics exceeds a threshold. Previous researchers have investigated the performance of the scan statistic to detect events; however, little attention was given to the optimization of which clusters the scan statistic should use. Using the scan …

A Novel Penalty-Based Wrapper Objective Function For Feature Selection In Big Data Using Cooperative Co-Evolution, A.N.M. Bazlur Rashid, Mohiuddin Ahmed, Leslie F. Sikos, Paul Haskell-Dowland

Research outputs 2014 to 2021

The rapid progress of modern technologies generates a massive amount of high-throughput data, called Big Data, which provides opportunities to find new insights using machine learning (ML) algorithms. Big Data consist of many features (also called attributes); however, not all these are necessary or relevant, and they may degrade the performance of ML algorithms. Feature selection (FS) is an essential preprocessing step to reduce the dimensionality of a dataset. Evolutionary algorithms (EAs) are widely used search algorithms for FS. Using classification accuracy as the objective function for FS, EAs, such as the cooperative co-evolutionary algorithm (CCEA), achieve higher accuracy, even …

Optimal Design Of A Flux Reversal Permanent Magnet Machine As A Wind Turbinegenerator, Majid Ghasemian, Farzad Tahami, Zahra Nasiri-Gheidari

Turkish Journal of Electrical Engineering and Computer Sciences

Flux reversal permanent magnet generators are well suited for use as wind turbine generators owing to their high torque generation ability and magnetic gear. However, they suffer from poor voltage regulation due to their high winding inductance. In this paper, a design optimization method is proposed for flux reversal generators in wind turbine applications. The proposed method includes a new multiobjective function. Cost, volume of the generator, and mass of the permanent magnet are considered in it independently and simultaneously. Besides the new objective function, the main superiority of this paper compared with published papers is considering winding inductance in …

Accurate Indoor Positioning With Ultra-Wide Band Sensors, Taner Arsan

Turkish Journal of Electrical Engineering and Computer Sciences

Ultra-wide band is one of the emerging indoor positioning technologies. In the application phase, accuracy and interference are important criteria of indoor positioning systems. Not only the method used in positioning, but also the algorithms used in improving the accuracy is a key factor. In this paper, we tried to eliminate the effects of off-set and noise in the data of the ultra-wide band sensor-based indoor positioning system. For this purpose, optimization algorithms and filters have been applied to the raw data, and the accuracy has been improved. A test bed with the dimensions of 7.35 m × 5.41 m …

Optimization Of Real-World Outdoor Campaign Allocations, Fatmanur Akdoğan Uzun, Doğan Altan, Ercan Peker, Mahmut Altuğ Üstün, Sanem Sariel

Turkish Journal of Electrical Engineering and Computer Sciences

In this paper, we investigate the outdoor campaign allocation problem (OCAP), which asks for the distribution of campaign items to billboards considering a number of constraints. In particular, for a metropolitan city with a large number of billboards, the problem becomes challenging. We propose a genetic algorithm-based method to allocate campaign items effectively, and we compare our results with those of nonlinear integer programming and greedy approaches. Real-world data sets are collected with the given constraints of the price class ratios of billboards located in İstanbul and the budgets of the given campaigns. The methods are evaluated in terms of …

Combined Analytic Hierarchy Process And Binary Particle Swarm Optimization Formultiobjective Plug-In Electric Vehicles Charging Coordination With Time-Of-Usetariff, Junaid Bin Fakhrul Islam, Mir Toufikur Rahman, Hazlie Mokhlis, Mohamadariff Othman, Tengku Fiaz Tengku Mohmed Noor Izam, Hasmaini Mohamad

Turkish Journal of Electrical Engineering and Computer Sciences

Plug-in electric vehicles (PEVs) are gaining popularity as an alternative vehicle in the past few years. The charging activities of PEVs impose extra electrical load on residential distribution system as well as increasing operational cost. There are multiple conflicting requirements and constraints during the charging activities. Therefore, this paper presents multiobjective PEV charging coordination based on weighted sum technique to provide simultaneous benefits to the power utilities and PEV users. The optimization problem of the proposed coordination is solved using binary particle swam optimization. The objectives of the coordination are to (i) minimize daily power loss, (ii) maximize power delivery …

Adaptive Modified Artificial Bee Colony Algorithms (Amabc) For Optimization Ofcomplex Systems, Rabi̇a Korkmaz Tan, Şebnem Bora

Turkish Journal of Electrical Engineering and Computer Sciences

Complex systems are large scale and involve numerous uncertainties, which means that such systems tend to be expensive to operate. Further, it is difficult to analyze systems of this kind in a real environment, and for this reason agent-based modeling and simulation techniques are used instead. Based on estimation methods, modeling and simulation techniques establish an output set against the existing input set. However, as the data set in a given complex systems becomes very large, it becomes impossible to use estimation methods to create the output set desired. Therefore, a new mechanism is needed to optimize data sets in …

Credible Optimum Selection Of Guidance System Simulation Based On Entropy Weight Vikor Method, Wenguang Yang, Yunjie Wu

Journal of System Simulation

Abstract: As the core component of the missile system, the guidance system plays an increasingly important role in the design of missile system. In order to improve the accuracy of the guidance system, this paper obtains the experimental data under a number of parameter design schemes by means of simulation experiments. The problem of simulation credibility verification of guidance system is transformed into multi-attribute decision-making optimization problem, and a parameter optimization method of guidance system based on improved VIKOR method is designed. The improved VIKOR method overcomes the phenomenon of rank reversal and ensures that the optimal final compromise solution …

Deep Representation Learning For Clustering And Domain Adaptation, Mohsen Kheirandishfard

Computer Science and Engineering Dissertations - Archive

Representation learning is a fundamental task in the area of machine learning which can significantly influence the performance of the algorithms used in various applications. The main goal of this task is to capture the relationships between the input data and learn feature representations that contain the most useful information of the original data. Such representations can be further leveraged in many machine learning applications such as clustering, natural language analysis, recommender systems, etc. In this dissertation, we first present a theoretical framework for solving a broad class of non-convex optimization problems. The proposed method is applicable to various tasks …

Interactive Fitness Domains In Competitive Coevolutionary Algorithm, Atm Golam Bari

USF Tampa Graduate Theses and Dissertations

Evolutionary Algorithms (EA) have been successfully applied to a wide range of optimization and search problems where no mathematical model of the quality of a candidate solution is available. Interactive Evolutionary Algorithms (IEA) and Competitive Coevolutionary Algorithms (CCoEA) go one step further by being able to tackle problems where the only means to evaluate the quality of a candidate solution is via interactions. In a typical IEA, interactions take place between the solution being evolved and human evaluators. In a CCoEA, interactions take place between solutions themselves, without need for human interaction. This dissertation identifies computer-aided learning as an application …

Optimal Sampling Paths For Autonomous Vehicles In Uncertain Ocean Flows, Andrew J. De Stefan

Dissertations

Despite an extensive history of oceanic observation, researchers have only begun to build a complete picture of oceanic currents. Sparsity of instrumentation has created the need to maximize the information extracted from every source of data in building this picture. Within the last few decades, autonomous vehicles, or AVs, have been employed as tools to aid in this research initiative. Unmanned and self-propelled, AVs are capable of spending weeks, if not months, exploring and monitoring the oceans. However, the quality of data acquired by these vehicles is highly dependent on the paths along which they collect their observational data. The …

Proton Exchange Membrane Fuel Cell Stack Design Optimization Using An Improved Jaya Algorithm, Uday K. Chakraborty

Educator Preparation & Leadership Faculty Works

Fuel cell stack configuration optimization is known to be a problem that, in addition to presenting engineering challenges, is computationally hard. This paper presents an improved computational heuristic for solving the problem. The problem addressed in this paper is one of constrained optimization, where the goal is to seek optimal (or near-optimal) values of (i) the number of proton exchange membrane fuel cells (PEMFCs) to be connected in series to form a group, (ii) the number of such groups to be connected in parallel, and (iii) the cell area, such that the PEMFC assembly delivers the rated voltage at the …

Improving Optimization Of Convolutional Neural Networks Through Parameter Fine-Tuning, Nicholas C. Becherer, John M. Pecarina, Scott L. Nykl, Kenneth M. Hopkinson

Faculty Publications

In recent years, convolutional neural networks have achieved state-of-the-art performance in a number of computer vision problems such as image classification. Prior research has shown that a transfer learning technique known as parameter fine-tuning wherein a network is pre-trained on a different dataset can boost the performance of these networks. However, the topic of identifying the best source dataset and learning strategy for a given target domain is largely unexplored. Thus, this research presents and evaluates various transfer learning methods for fine-grained image classification as well as the effect on ensemble networks. The results clearly demonstrate the effectiveness of parameter …

Some Theoretical Links Between Shortest Path Filters And Minimum Spanning Tree Filters, Sravan Danda, Aditya Challa, B. S.Daya Sagar, Laurent Najman

Journal Articles

Edge-aware filtering is an important pre-processing step in many computer vision applications. In the literature, there exist several versions of collaborative edge-aware filters based on spanning trees and shortest path heuristics which work well in practice. For instance, tree filter (TF) which is recently proposed based on a minimum spanning tree (MST) heuristic yields promising results in many filtering applications. However, links between the tree-based filters and shortest path-based filters are faintly explored. In this article, we introduce an edge-aware generalization of the TF termed as UMST filter based on a subgraph generated by edges of all MSTs. The major …

Re-Org: An Online Repositioning Guidance Agent, Muralidhar Konda, Pradeep Varakantham, Aayush Saxena, Meghna Lowalekar

Research Collection School Of Computing and Information Systems

No abstract provided.

Learning Two-Layer Neural Networks With Symmetric Inputs, Rong Ge, Rohith Kuditipudi, Zhize Li, Xiang Wang

Research Collection School Of Computing and Information Systems

We give a new algorithm for learning a two-layer neural network under a very general class of input distributions. Assuming there is a ground-truth two-layer network $y = A \sigma(Wx) + \xi$, where A, W are weight matrices, $\xi$ represents noise, and the number of neurons in the hidden layer is no larger than the input or output, our algorithm is guaranteed to recover the parameters A, W of the ground-truth network. The only requirement on the input x is that it is symmetric, which still allows highly complicated and structured input. Our algorithm is based on the method-of-moments framework …

Enhancing Portability In High Performance Computing: Designing Fast Scientific Code With Longevity, Jason Orender

Computer Science Theses & Dissertations

Portability, an oftentimes sought-after goal in scientific applications, confers a number of possible advantages onto computer code. Portable code will often have greater longevity, enjoy a broader ecosystem, appeal to a wider variety of application developers, and by definition will run on more systems than its pigeonholed counterpart. These advantages come at a cost, however, and a rational approach to balancing costs and benefits requires a systemic evaluation. While the benefits for each application are likely situation-dependent, the costs in terms of resources, including but not limited to time, money, computational power, and memory requirements, are quantifiable. This document will …

Two-On-One Pursuit With A Non-Zero Capture Radius, Patrick J. Wasz

Theses and Dissertations

In this paper, we revisit the "Two Cutters and Fugitive Ship" differential game that was addressed by Isaacs, but move away from point capture. We consider a two-on-one pursuit-evasion differential game with simple motion and pursuers endowed with circular capture sets of radius l > 0. The regions in the state space where only one pursuer effects the capture and the region in the state space where both pursuers cooperatively and isochronously capture the evader are characterized, thus solving the Game of Kind. Concerning the Game of Degree, the algorithm for the synthesis of the optimal state feedback strategies of the …

Centroidal Voronoi Tessellation With Local Optimization, Tianyu Ye, Yiqun Wang, Dongming Yan, Junhai Yong

Journal of System Simulation

Abstract: Centroidal Voronoi tessellation is a special geometric structure, which has many applications in various fields such as geographical information system, signal processing, mesh generation/optimization, visualization and so on. Due to the highly non-convex nature of the CVT energy function, the existing methods for computing CVT have several drawbacks, which always trap into local minima. We propose generation optimization and stochastic optimization schemes for further reducing the CVT energy. Experimental results show that the proposed method improves both quality and efficiency compared to the recent approaches.

An Evaluation Of Training Size Impact On Validation Accuracy For Optimized Convolutional Neural Networks, Jostein Barry-Straume, Adam Tschannen, Daniel W. Engels, Edward Fine

SMU Data Science Review

In this paper, we present an evaluation of training size impact on validation accuracy for an optimized Convolutional Neural Network (CNN). CNNs are currently the state-of-the-art architecture for object classification tasks. We used Amazon’s machine learning ecosystem to train and test 648 models to find the optimal hyperparameters with which to apply a CNN towards the Fashion-MNIST (Mixed National Institute of Standards and Technology) dataset. We were able to realize a validation accuracy of 90% by using only 40% of the original data. We found that hidden layers appear to have had zero impact on validation accuracy, whereas the neural …

Artificial Fish Swarm And Feedback Linearization Of Flue Gas Denitration Control Based On Neural Network, Yuguang Niu, Pan Yan, Wenyuan Huang

Journal of System Simulation

Abstract: According to the present situation of SCR flue gas dentration control system in thermal power plant, an optimum proposal that control valve and concentration transmitter are added in the inlet of the SCR reactor is presented, and the corresponding control strategy is given. At the entrance of the SCR reactor, the receding horizon algorithm combined with the single neuron adaptive algorithm and the artificial fish swarm algorithm (RSNAAFS) is used to control branch valves to pretreat NO_X in the exhaust flue gas. At the outlet of the SCR reactor, the neural network based on feedback linearization algorithm (NNFL) …

Real-Time Pricing Strategy Considering The Risk Of Smart Grid, Hongbo Zhu, Gao Yan, Yeming Dai

Journal of System Simulation

Abstract: The real-time electricity price mechanism is an ideal method to adjust the power balance between supply and demand in smart grid. Its implementation has profound impacts on the users' behavior and the operation and management of electricity power grid’s safety. The users’ demand behavior plays a regulatory role in designing real-time electricity pricing strategy. Aiming at maximizing social welfare, the dynamic change of users’ aggregate demand is analyzed, which corrects the electricity risk items in online real-time risk model in the way of changing the individual user’s power fluctuations to all the users’ demand power fluctuations, and the optimization …

Optimization Of Q-Btgsid Based On Sensitivity Analysis, Shuwei Jia

Journal of System Simulation

Abstract: In order to make up for the defect that relative degree of incidence, absolute degree of incidence and synthetic degree of incidence are limited in the range of (0.5, 1], this paper attempts to improve the degree of grey incidence. A control factor of “λ” and the metric space are set up to adjust. A new model is established and its specific properties are studied. It is proved that the new model satisfies the grey incidence axioms and the range of degree of grey incidence can be extended to (0, 1]. We put forward four principles of …

Optimization Of Scheduling Rule Of Unidirectional Material Handling System With Short-Cut, Juntao Li, Kun Xia, Kise Hiroshi

Journal of System Simulation

Abstract: To decrease the interference and improve the performance of a unidirectional circulation-type material handling system on a single loop with a shortcut, the interference and scheduling problem between AGVs are studied. According to the actual situation of material handling system, the interferences of two scheduling rules (random rule and order rule) are analyzed. An optimal scheduling rule under the interference case—exchange order rule is proposed. Different scheduling rules have an influence on the interference between AGVs and then have an important effect on the efficiency of the whole system. Experiment results show that the exchange order (E-Order) rule …

Performance Enhancement Of Photovoltaic System Using Genetic Algorithm- Based Maximum Power Point Tracking, Brammanayagam Nagarani, Jothiswaroopan Nesamony

Turkish Journal of Electrical Engineering and Computer Sciences

In recent years, enormous progress has been made on power generation using photovoltaic (PV) system. Solar power is one of the most promising renewable energy sources that is providing its benefit specifically in rural areas. With the increasing need for solar energy, it becomes necessary to extract maximum power from the PV array. The output power of the solar cells varies directly with the ambient temperature and Irradiation. Therefore, the challenge is to track maximum power from the PV array when environmental factors change. This paper focuses on increasing the efficiency of a PV array by incorporating artificial intelligence techniques. …