Summer School on Computational Intelligence and Applications (SSoCIA)-2018


2018 IEEE CIS Summer School on Computational Intelligence and Applications (2018 IEEE CIS SSoCIA)

 

Introduction

On this year, the Summer School in CI and Applications (2018 IEEE SSoCIA) will be held in Guadalajara-Mexico, co-located with IEEE LA-CCI 2018. Its purpose is to provide to students, professors, researchers and professionals who are interested in Computer Science and Computational Intelligence to profit from the expertise of the speakers of the main conference. The short courses offered are thought to be useful to both, scientific research as well as application, addressing several problems in industry, commerce, and government.

This Summer School will be held immediately during the LA-CCI, which will allow complementary objectives of (1) to strengthen fundamental knowledge in CI during the Summer School, and (2) to update the state of the art understanding in CI during the main conference (November 7-9th).

 

Objectives

The objective of this Summer School is to provide a platform for Latin-American young researchers and students to experience the recent developments, get hold of the state-of-the-art in computational intelligence methods and applications in several areas. The main goal is to provide to young researchers the opportunity to interact with experts and eminences in CI and exchange ideas and experiences.

In addition, the summer school provides demonstration sessions on CI method, where the participants can obtain experience in situ for practical applications in the topic of interest of 2018 IEEE Latin American Conference on Computational Intelligence (2018 IEEE LA-CCI).

 

*** Program (WORKSHOPS/TUTORIALS) – OVERVIEW on November 6, 2018:

  1. GOOGLE’S WORKSHOP, Machine Learning at Google, Dr Fabiel Zuniga
  2. INTEL´S WORKSHOP (A), Introduction to Machine Learning with Matlab’s Neural Networks toolbox, Héctor Alfonso Cordourier and Dr. Paulo Lopez, INTEL
  3. INTEL´S WORKSHOP (B), Higher order neural networks, Dr. Julio Zamora, INTEL
  4. GENERAL WORKSHOP (A): Learning navigation policies in unstructured terrain using intelligent systems (In Spanish), Dr. Roberto Valencia-Murillo
  5. GENERAL WORKSHOP (B): Planeación de trayectorias para un robot móvil basada en algoritmos de optimización metaheurísticos (In Spanish), Prof. José de Jesús Hernández Barragán
  6. GENERAL WORKSHOP (C): Germinal Center Optimization Algorithm for Robotic mapping (In Spanish), Dr. Carlos Villaseñor

 

*** Program (WORKSHOPS/TUTORIALS) – DESCRIPTIONS:

1. GOOGLE’S WORKSHOP: Machine Learning at Google (Fabiel Zuniga, GOOGLE)

Abstract: Machine Learning at Google — what is it, how are we staying at the forefront of ML research, and where are we applying ML to our products at Google. 2 HOURS

Biography: Graduated in Computer Engineering in 2000 from the Center of Exact Sciences and Engineering (CUCEI, Guadalajara, Jal. Mex. – www.cucei.udg.mx). After college he worked for Petroleos Mexicanos (PEMEX) for 8 months in the Systems Department where he developed office automation tools. Shortly after, he pursued his Master’s degree in Computer Science at the Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV, Guadalajara, Jal. Mex. – www.gdl.cinvestav.mx), and graduated in 2002. Five years later, he was awarded a PhD in Computer Science from the same institution in 2007. During the final stage of his PhD he worked at ASCI, a Mexican company that among other projects, outsourced projects to Hewlett-Packard. In 2088 he got an offer from Hewlett-Packard USA, where he worked on the ProCurve division on Network Management tools until 2015 – filing 4 patents with the company. Since then, he’s been working at the Google headquarters in Mountain View, CA on Geo Merchant Platforms for Google maps.

 

2. INTEL’S WORKSHOPS (A): Introduction to Machine Learning with Matlab’s Neural Networks toolbox (Dr. Héctor Alfonso Cordourier and Dr. Paulo Lopez, INTEL)

Abstract: This workshop is intended to provide a quick, high level introduction to one of the most popular and promising Machine Learning techniques: Neural Networks; using Matlab’s Neural Networks toolbox. We will perform actual trainings and tests of neural nets using real-life data in different applications, with the objective of understanding the real potential of this kind of tools to solve nowadays problems. 1 HOUR

Biography: Dr. Héctor Alfonso Cordourier, was born in Mexico City on December 18, 1980. He completed his undergraduate studies in Physical Engineering at the Autonomous University of Yucatan, where he studied instrumentation and signal processing. Afterwards, he completed his graduate studies in Electrical Engineering at the National Autonomous University of Mexico, where he specialized in the field of Acoustic Instrumentation and performed studies of musical instruments and active noise control. He currently works at Intel in Guadalajara, where he develops projects related to acoustics, speech recognition, improvement of user experience, and Artificial Intelligence applied to voice processing. Paulo Lopez Meyer received the B.E. degree in Telecommunications Engineering in 2003, and the M.E. degree in Instrumentation Engineering in 2005, both from the National Autonomous University of Mexico, in Mexico City. He received the Ph.D. degree in Electrical and Computer Engineering from Clarkson University, located in Potsdam, New York, in 2010, where he worked in the development of several biomedical sensors. From 2010 through 2012, he worked as a Post-doctoral fellow in The University of Alabama, in Tuscaloosa, Alabama, continuing his research in the biomedical field. Early in 2013, he started working for Intel Labs in Guadalajara, Mexico, as an Algorithm Engineer performing R&D focused on portable and wearable sensor technology, applying methodologies related to artificial intelligence and machine learning. As part of his academic career, Paulo has been part of the SNI from CONACyT, he has contributed to 40+ articles published in several conferences and journals of international renown, and has participated in the development of 25 patent applications, of which 4 have been already granted in 2018. His research interests include the application of data analysis, signal processing, machine learning and pattern recognition techniques in the solution of real-life problems.

 

3. INTEL’S WORKSHOPS (B): Higher order neural networks (Dr. Julio Zamora, INTEL)

Abstract: An overview of higher order neural networks. By increasing the order of the NN, the number of layers decreases, resulting in a reduction of the number of activation functions required, and reducing the time consumed for training. Higher order neural networks do not introduce more transcendental functions; instead, it tries to reduce the need of them by reducing the need of layers and for each layer removed, we can reduce hundreds of activation functions. 1 HOUR

Biography: Julio Zamora received his Bachelor’s degree as an Electronics Engineer in 2000 at the ITCG University, in Mexico.  In 2003, he completed his M.S. in computer sciences doing research related to movile robots. In 2006, he completed his Ph.D. studies at CINVESTAV, in Mexico. The core research topic for his thesis was the study of robotic object manipulation guided by computer vision.  In 2007, Julio had a postdoctoral position at KAIST, Korea, working with computer vision algorithms for humanoid robots at Hubo lab, Julio joins Intel in 2007 for R&D as research scientist, and he is author or co-author of 19 US patents in process. Julio’s research interests include Artificial intelligence, Computer vision, geometric algebras, Robotics, Image processing. Julio Zamora was nominated for the W.K. Clifford international price for his contributions to geometric algebra applications. Member of the IEEE, the SNI-C, and the MACVNR, and a recurrent professor of the Masters of Engineering program at the ITESM University in Guadalajara, Mexico.

 

4. GENERAL WORKSHOP (A): Learning navigation policies in unstructured terrain using intelligent systems – In Spanish (Dr. Roberto Valencia-Murillo)

Abstract: In this talk, the complex task of learning navigation policies in an unstructured terrain is going to be addressed using intelligent systems. First, the learning to search (LEARCH) algorithm is going to be explained, LEARCH constructs cost functions that map environmental features to a certain cost for traversing a patch of terrain. These features are abstractions of the environment, in which trees, vegetation, slopes, water, and rocks can be found, and the traversal costs are scalar values that represent the difficulty for a robot to cross given patches of terrain. Using these costs, a robot can plan an optimal path that avoids dangerous terrain. However, LEARCH tends to forget knowledge after new policies are learned. So, as a second part of the talk, the use of reinforcement learning and long–short term memory neural networks (LSTM) to provide a memory for LEARCH is explained. This approach allows the knowledge learned in previous training to be used to navigate new environments and, also, for retraining.

Requirements: Any C ++ compiler can be used, I’ve been working on several versions of my code, it works on Linux, and I’m preparing it to work with Visual Studio, in case you want to work with Windows. 2 HOURS

Biography: Roberto Valencia-Murillo has received the Ph.D. degree in Computer Science Department, at the Centro Universitario de Ciencias Exactas e Ingeniería at Universidad de Guadalajara in México in 2017. He has received a Master’s degree in Computer Science in 2013 and a B.Sc. degree in Computer Science at the University of Baja California. His research interests focus in machine learning, robot navigation and geometric methods for robotics

 

5. GENERAL WORKSHOP (B): Planeación de trayectorias para un robot móvil basada en algoritmos de optimización metaheurísticos – In Spanish (Prof. José de Jesús Hernández Barragán)

Resumen: Una tarea esencial para robots móviles es moverse a través de su entorno de forma segura. Por lo tanto, la selección de un algoritmo confiable para la determinación de trayectorias es un punto clave. En este taller, los autores presentan un método para resolver la planeación de trayectorias como un problema de optimización global con restricciones. Para resolver el problema de optimización, se considera utilizar los algoritmos metaheurísticos de Optimización por Enjambre de Partículas y Evolución Diferencial. Además, se propone utilizar funciones de penalización para incluir las restricciones en el proceso de optimización. Finalmente, el método propuesto se enfoca en resolver problemas de planeación de trayectorias global. Sin embargo, este método puede ser extendido para la planeación de trayectorias local.

Materiales: Se pretende compartir algoritmos programados en MATLAB, aunque estos son compatibles con OCTAVE. 2 HOURS

Biography: Jose de Jesus Hernandez-Barragan, is a Ph.D. student in the Computer Sciences Department, in the Centro Universitario de Ciencias Exactas e Ingeniería at the Universidad de Guadalajara in México. He received a Master’s degree in Computer Science in 2015 and a B.Sc. degree as Computer Engineer at the same university. His research focuses in bio-inspired algorithms and applications to robotics and vision.

 

6. GENERAL WORKSHOP (C): Germinal Center Optimization Algorithm for Robotic mapping – In Spanish (Dr. Carlos Villaseñor)

Abstract: The vertebrate’s immune system has the difficult task of protecting the body from foreign substances. The immune system shows a high ability of adaptation and spatialization for detecting antigens (Ag), we call this process Affinity Maturation (AM).  The highest AM is obtained in a process called Germinal Center Reaction. In this course, we present a brief introduction to the Germinal Center Optimization (GCO) Algorithm which is a multivariate optimization technique based in the Germinal Center reaction. GCO algorithm implements a competitive-based non-uniform distribution to select particles to be mutated. The GCO is capable of behavior like different algorithms through the automatic adaptation of the leadership. We also show an application of ellipsoidal mapping using Geometric Algebra.

Course contents 2 HOURS:

  • Biological phenomenon explained
  • Algorithm explanation
  • Code explanation (the code will be released for assistants)
  • Algorithm benchmarking
  • Review of some already publish applications (Neuro-identification and Neuro-control)
  • Tour for ellipsoidal mapping algorithm

Requirements:

  • We will provide two versions of the algorithm in C++ and Python 3

-For C++, the code in project in Visual Studio 2017 with standard libraries will be provided.

-For Python 3, and implementation using NumPy will be released.

Biography: Carlos Villaseñor has received the Ph.D. degree in Computer Science Department, at the Centro Universitario de Ciencias Exactas e Ingeniería at Universidad de Guadalajara in México in 2017. He received a Master’s degree in Computer Science in 2014 and a B.Sc. degree in Industrial Engineer at the same university. His research focuses in bio-inspired algorithms and applications of geometric algebras.

 

IEEE LA-CCI School Directors: Carlos Lopez-Franco and Nancy Arana-Daniel, Mexico