Hyper Parameter Optimization, seminário com Dr Everton Gomede
Título: Hyper Parameter Optimization, Automated Machine Learning, Gaussian Process, Artificial Neural Network, Bayesian Optimization, com o Dr. Everton Gomede
Transmissão: https://youtu.be/pMt-JPFf0Ig
Resumo: Hyper parametrization is a usual task in machine learning. It aims to build different models and select the best one. However, selecting the best combinations of hyper parameters is a challenge. This happens because the number of combinations resulting from values and hyper parameters available is enormous. Also, it is needed to deal with three types of hyper parameters, discrete, continuous, and conditionals. Thus, the combinatorial result implies a NP-hard problem.
In that context, we present a Bayesian Optimization of hyper parameters set aims to find the extrema of loss function. The results point out that it is possible to identify suitable sets of hyper parameters that could be solved in polynomial time. This approach might be applied as a solution for hyper parametrization tasks in general experiments and/or AutoML.
Bio: Everton Gomede is Computer Scientist at VanHack (Canada), at FEEC/Unicamp (Brazil) where we develop algorithmic processes, solutions, and tools that enable these companies and its analysts to efficiently extract insights from data and provide solution alternatives to decision-makers.Work on applied analytic research, develop data analysis software, create data science educational content, write books, and provide analytic consulting services. Besides, I have been working as a Visiting Professor at the University of São Paulo, University of Maringá, Pontificia Universidad Catolica, FCV, and University of Londrina where I have taught business analytics and R and Python programming courses. My teaching style is building knowledge from general to specific, simple to complex and example-based methods before mathematical formalization.
Also reviewer of journals Journal of Waste Resources and Recycling (JWRR), Trends in Computer Science and Information Technology, Asian Journal of Sociological Research, Asian Journal of Education and Social Studies, Trends in Computer Science and Information Technology, South Asian Journal of Social Studies and Economics, Artificial Intelligence Review, Machine Learning Research, ISPRS International Journal of Trends in Computer Science and Information Technology, Geo-Information, Smart Cities, Sensors, International Journal of Environmental Research and Public Health, Applied Sciences, Big Data and Cognitive Computing, and Future Internet. Moreover, I have written many technical books and scientific papers related to computer science and I have been working as a researcher at FEEC/Unicamp.
