Demonstrating the interplay of machine learning and optimization methods for operational planning decision

Rapid growth of network traffic causes the need for the development of new network technologies. Artificial intelligence provides suitable tools to improve currently used network optimization methods. In this paper, we propose a procedure for network traffic prediction. Based on optical networks’ (and other network technologies) characteristics, we focus on the prediction of fixed bitrate levels called traffic levels. We develop and evaluate two approaches based on different supervised machine learning (ML) methods—classification and regression. We examine four different ML models with various selected features. The tested datasets are based on real traffic patterns provided by the Seattle Internet Exchange Point (SIX). Obtained results are analyzed using a new quality metric, which allows researchers to find the best forecasting algorithm in terms of network resources usage and operational costs. Our research shows that regression provides better results than classification in case of all analyzed datasets. Additionally, the final choice of the most appropriate ML algorithm and model should depend on the network operator expectations.

Download Full-text

Machine Learning for Design Optimization of Electromagnetic Devices: Recent Developments and Future Directions

Applied Sciences ◽

10.3390/app11041627 ◽

2021 ◽

Vol 11 (4) ◽

pp. 1627

Author(s):

Yanbin Li ◽

Gang Lei ◽

Gerd Bramerdorfer ◽

Sheng Peng ◽

Xiaodong Sun ◽

...

Keyword(s):

Machine Learning ◽

Design Optimization ◽

Optimization Methods ◽

Machine Learning Algorithms ◽

Cloud Services ◽

Robust Design Optimization ◽

Support Vector ◽

Future Directions ◽

Electromagnetic Devices ◽

Recent Developments

This paper reviews the recent developments of design optimization methods for electromagnetic devices, with a focus on machine learning methods. First, the recent advances in multi-objective, multidisciplinary, multilevel, topology, fuzzy, and robust design optimization of electromagnetic devices are overviewed. Second, a review is presented to the performance prediction and design optimization of electromagnetic devices based on the machine learning algorithms, including artificial neural network, support vector machine, extreme learning machine, random forest, and deep learning. Last, to meet modern requirements of high manufacturing/production quality and lifetime reliability, several promising topics, including the application of cloud services and digital twin, are discussed as future directions for design optimization of electromagnetic devices.

Download Full-text

Reviewing the application of machine learning methods to model urban form indicators in planning decision support systems : potential, issues and challenges

Journal of King Saud University - Computer and Information Sciences ◽

10.1016/j.jksuci.2021.08.007 ◽

2021 ◽

Author(s):

Stephane Cedric KOUMETIO TEKOUABOU ◽

El Bachir Diop ◽

Rida Azmi ◽

Remi Jaligot ◽

Jerome Chenal

Keyword(s):

Machine Learning ◽

Decision Support ◽

Decision Support Systems ◽

Urban Form ◽

Support Systems ◽

Learning Methods ◽

Machine Learning Methods ◽

Planning Decision

Download Full-text

Identifying cognitive deficits in cocaine dependence using standard tests and machine learning

10.31234/osf.io/7qsz3 ◽

2018 ◽

Author(s):

Said Enrique Jiménez ◽

Diego Angeles-Valdez ◽

Viviana Villicaña ◽

Ernesto Reyes-Zamorano ◽

ruth alcala ◽

...

Keyword(s):

Machine Learning ◽

Cocaine Dependence ◽

Diagnosis And Treatment ◽

Cognitive Domains ◽

Cognitive Markers ◽

Multiple Observations ◽

Other Substances ◽

Planning Decision ◽

The Mind ◽

Treatment Improvement

Background and aims There is a growing need for detecting valid and generalizable markers due to a demand of accurate Cocaine Dependence diagnosis and treatment. Machine Learning (ML) is a modern statistical alternative to select from multiple observations the most reliable features, which allows precise and more effective categorization addressing the demand to improve diagnosis. The aim of the current study was to identify cognitive markers by using three ML algorithms, Elastic Net (GlmNet), Random forest (Rf) and Generalized Linear Model (Glm), with the purpose of classify Cocaine Dependence (CD) and Non-dependent controls (NDC) to make it generalizable for new samples. Methods Two independent samples were required, the first one consisted on 87 participants (53 CD and 34 NDC) and the second one conformed by 40 participants (20 CD and 20 NDC). All participants were evaluated with neuropsychological tests that included 40 variables assessing cognitive domains of flexibility, inhibition, working memory, problem solving, planning, decision making and theory of the mind. With the results of the cognitive evaluation the three ML algorithms were trained in the first sample and tested on the second one to classify into CD and NDC. Results Even though the three algorithms had a ROC performance over 50%, GlmNet was superior in both, training (ROC = 0.71) and testing set (ROC = 0.85) compared to Rf and Glm. Furthermore, GlmNet was capable of identifying eight predictors out of 40 from all the cognitive domains assessed. Conclusions ML is an effective approach for the identification of generalizable cognitive markers. Specific subsets resulted robust predictors for accurate classification of new cases, such as those from cognitive flexibility and inhibition domain. These findings are relevant in addictions field as they have highly beneficial potential for diagnosis and treatment improvement, not only for CD but also for other substances abuse.

Download Full-text

Machine Learning Enhanced Spectroscopic Analysis: Towards Autonomous Chemical Mixture Characterization for Rapid Process Optimization

Digital Discovery ◽

10.1039/d1dd00027f ◽

2022 ◽

Author(s):

Andrea Angulo ◽

Lankun Yang ◽

Eray S Aydil ◽

Miguel A. Modestino

Keyword(s):

Machine Learning ◽

Process Optimization ◽

Parameter Space ◽

Chemical Process ◽

Spectroscopic Analysis ◽

Process Development ◽

Operating Parameter ◽

Optimization Methods ◽

Chemical Mixture ◽

Mixture Characterization

Autonomous chemical process development and optimization methods use algorithms to explore the operating parameter space based on feedback from experimentally determined exit stream compositions. Measuring the compositions of multicomponent streams...

Download Full-text

Applying machine learning optimization methods to the production of a quantum gas

Machine Learning: Science and Technology ◽

10.1088/2632-2153/ab6432 ◽

2020 ◽

Vol 1 (1) ◽

pp. 015007 ◽

Cited By ~ 4

Author(s):

A J Barker ◽

H Style ◽

K Luksch ◽

S Sunami ◽

D Garrick ◽

...

Keyword(s):

Machine Learning ◽

Optimization Methods ◽

Quantum Gas

Download Full-text

Evaluating Optimization Strategies for Engine Simulations Using Machine Learning Emulators

Journal of Engineering for Gas Turbines and Power ◽

10.1115/1.4043964 ◽

2019 ◽

Vol 141 (9) ◽

Author(s):

Daniel M. Probst ◽

Mandhapati Raju ◽

Peter K. Senecal ◽

Janardhan Kodavasal ◽

Pinaki Pal ◽

...

Keyword(s):

Machine Learning ◽

Random Search ◽

Optimization Algorithms ◽

Optimization Methods ◽

Machine Learning Algorithms ◽

Cfd Simulations ◽

Engine Combustion ◽

Computational Fluid Dynamics Cfd ◽

Micro Genetic Algorithm ◽

Time Required

This work evaluates different optimization algorithms for computational fluid dynamics (CFD) simulations of engine combustion. Due to the computational expense of CFD simulations, emulators built with machine learning algorithms were used as surrogates for the optimizers. Two types of emulators were used: a Gaussian process (GP) and a weighted variety of machine learning methods called SuperLearner (SL). The emulators were trained using a dataset of 2048 CFD simulations that were run concurrently on a supercomputer. The design of experiments (DOE) for the CFD runs was obtained by perturbing nine input parameters using a Monte-Carlo method. The CFD simulations were of a heavy duty engine running with a low octane gasoline-like fuel at a partially premixed compression ignition mode. Ten optimization algorithms were tested, including types typically used in research applications. Each optimizer was allowed 800 function evaluations and was randomly tested 100 times. The optimizers were evaluated for the median, minimum, and maximum merits obtained in the 100 attempts. Some optimizers required more sequential evaluations, thereby resulting in longer wall clock times to reach an optimum. The best performing optimization methods were particle swarm optimization (PSO), differential evolution (DE), GENOUD (an evolutionary algorithm), and micro-genetic algorithm (GA). These methods found a high median optimum as well as a reasonable minimum optimum of the 100 trials. Moreover, all of these methods were able to operate with less than 100 successive iterations, which reduced the wall clock time required in practice. Two methods were found to be effective but required a much larger number of successive iterations: the DIRECT and MALSCHAINS algorithms. A random search method that completed in a single iteration performed poorly in finding optimum designs but was included to illustrate the limitation of highly concurrent search methods. The last three methods, Nelder–Mead, bound optimization by quadratic approximation (BOBYQA), and constrained optimization by linear approximation (COBYLA), did not perform as well.

Download Full-text

Artificial intelligence applied to small businesses: the use of automatic feature engineering and machine learning for more accurate planning

Revista de Contabilidade e Organizações ◽

10.11606/issn.1982-6486.rco.2020.171481 ◽

2020 ◽

Vol 14 ◽

pp. e171481

Author(s):

Alexandre Moreira Nascimento ◽

Vinicius Veloso De Melo ◽

Anna Carolina Muller Queiroz ◽

Thomas Brashear-Alejandro ◽

Fernando de Souza Meirelles

Keyword(s):

Machine Learning ◽

Small Businesses ◽

Operational Planning ◽

Feature Engineering ◽

The Novel ◽

Rule Based ◽

Novel Technique ◽

Using Data ◽

Rule Based Classifier ◽

Medium Businesses

The purpose of this study is to develop a predictive model that increases the accuracy of business operational planning using data from a small business. By using Machine Learning (ML) techniques feature expansion, resampling, and combination techniques, it was possible to address several existing limitations in the available research. Then, the use of the novel technique of feature engineering allowed us to increase the accuracy of the model by finding 10 new features derived from the original ones and constructed automatically through the nonlinear relationships found between them. Finally, we built a rule-based classifier to predict the store's revenue with high accuracy. The results show the proposed approach open new possibilities for ML research applied to small and medium businesses.

Download Full-text

Overparameterized neural networks implement associative memory

Proceedings of the National Academy of Sciences ◽

10.1073/pnas.2005013117 ◽

2020 ◽

Vol 117 (44) ◽

pp. 27162-27170

Author(s):

Adityanarayanan Radhakrishnan ◽

Mikhail Belkin ◽

Caroline Uhler

Keyword(s):

Machine Learning ◽

Neural Networks ◽

Associative Memory ◽

Deep Neural Networks ◽

Optimization Methods ◽

Efficient Mechanism ◽

Training Samples ◽

Sample Recovery ◽

Sequence Encoding ◽

Standing Problem

Identifying computational mechanisms for memorization and retrieval of data is a long-standing problem at the intersection of machine learning and neuroscience. Our main finding is that standard overparameterized deep neural networks trained using standard optimization methods implement such a mechanism for real-valued data. We provide empirical evidence that 1) overparameterized autoencoders store training samples as attractors and thus iterating the learned map leads to sample recovery, and that 2) the same mechanism allows for encoding sequences of examples and serves as an even more efficient mechanism for memory than autoencoding. Theoretically, we prove that when trained on a single example, autoencoders store the example as an attractor. Lastly, by treating a sequence encoder as a composition of maps, we prove that sequence encoding provides a more efficient mechanism for memory than autoencoding.

Download Full-text

Chemical Robotics Enabled Exploration of Stability and Photoluminescent Behavior in Multicomponent Hybrid Perovskites via Machine Learning

10.26434/chemrxiv.12436079.v1 ◽

2020 ◽

Author(s):

Kate Higgins ◽

Sai Mani Valleti ◽

Maxim Ziatdinov ◽

Sergei Kalinin ◽

Mahshid Ahmadi

Keyword(s):

Machine Learning ◽

Optimization Methods ◽

Combinatorial Synthesis ◽

Ambient Conditions ◽

Term Stability ◽

Long Term Stability ◽

Composition Space ◽

Materials Research ◽

The Stability

Hybrid organic-inorganic perovskites have attracted immense interest as a promising material for the next-generation solar cells; however, issues regarding long-term stability still require further study. Here, we develop automated experimental workflow based on combinatorial synthesis and rapid throughput characterization to explore long-term stability of these materials in ambient conditions, and apply it to four model perovskite systems: MAxFAyCs1-x-yPbBr3, MAxFAyCs1-x-yPbI3, (CsxFAyMA1-x-yPb(Brx+yI1-x-y)3) and (CsxMAyFA1-x-yPb(Ix+yBr1-x-y)3). We also develop a machine learning-based workflow to quantify the evolution of each system as a function of composition based on overall changes in photoluminescence spectra, as well as specific peak positions and intensities. We find the stability dependence on composition to be extremely non-uniform within the composition space, suggesting the presence of potential preferential compositional regions. This proposed workflow is universal and can be applied to other perovskite systems and solution-processable materials. Furthermore, incorporation of experimental optimization methods, e.g., those based on Gaussian Processes, will enable the transition from combinatorial synthesis to guide materials research and optimization.

Download Full-text