24 Oct, 2017 · 11:30 al 24 Oct, 2017 · 13:00 CITIC-UGR
Conferencia: “Demystifying Reservoir Computing”
Ali Rodan
Conferencias, seminarios, divulgación científica
tweet facebook
- Fecha: Martes, 24 de octubre de 2017
- Lugar: Sala de Conferencias (CITIC-UGR)
- Horario: 11:30h
- Descripción: Author: Dr. Ali Rodan. Associate Professor, The University of Jordan.
- Abstract: In this talk Dr. Ali Rodan will present one of the simple reservoir computing methods called Echo State Network (ESN) which is a type of recurrent neural network that has a fixed random input and hidden layers and a trainable output layer, and ask:
- What is the minimal complexity of reservoir construction for obtaining competitive models?
- Can the introducing of jumps (shortcuts) in the network hidden layer for simple reservoir construction improve performance results?
- Can we get superior results when performing reservoir computing models on different applications including Credit Risk Evaluation, Churn Prediction, and Manufacturing Process Modeling?
- Can the use of diverse ESN reservoirs whose collective readout is obtained through Negative Correlation Learning (NCL) achieve better generalization performance?
- Speaker: Dr. Ali Rodan (M.Sc. Oxford Brookes University 2005; Ph.D. The University of Birmingham 2012) is an Associate Professor of Computer science at the University of Jordan, Amman, Jordan. His research interests include recurrent neural networks, dynamical systems and machine learning. Currently he specializes on reservoir computation models including their design and theoretical properties.
- Abstract: In this talk Dr. Ali Rodan will present one of the simple reservoir computing methods called Echo State Network (ESN) which is a type of recurrent neural network that has a fixed random input and hidden layers and a trainable output layer, and ask:
- Organiza: ETS Ingenierías Informática y de Telecomunicación. La visita de investigación del profesor Rodan ha sido financiada por el programa Erasmus Plus Faculty Exchange.
- Más información: pacv@ugr.es
