Lehrstuhl Kognitive Integrierte Sensorsysteme: Neurocomputing

Neurocomputing (V2/Ü1)

Instructor: Prof. Dr.-Ing. Andreas König

Lab/Exercises: Prof. Dr.-Ing. Andreas König and research assistants

Consultations: During the lecture period every Monday, 13:00-14:00, 12/455, else according to bulletin announcement or personal appointment (koenig@eit.uni-kl.de)

Contents:

Qualification aims:

Understanding of the concepts of dedicated neural and bio-inspired hardware and its application potential and limitations
Understanding of design principles of circuits with alternative signal representation and adaptive structures
Understanding of the effect of simplified implementations
Understanding of fault-tolerance and robustness issues
Ability of contents/result abstraction to MEMS/microsystems applications

Teaching contents:

Introduction to the field of innovative computer architectures and systems for the technical implementation of biological information processing principles
Presentation of diverse aims and solution concepts: Hardware for technical cognition systems, biological-technical interfaces, simulation and verification of models of biological evidence
Rehearsal of relevant and commonly applied neural algoriths and analysis of computational requirements and operators
Presentation and effect of potential simplifcation options for the regarded algorithms
Basics of circuit technology (digital, analog, opto-elektronic/optisch) and related implementation technologies (CMOS, WSI, MEMS, etc.) for neural hardware
Overview of fundamental architectural principles of neurochips, -processors and -computers
Assessment criteria and taxonomy for neural HW
Presentation and detailed discussion of selected, representative implementations
Outlook on new lines in the field, e.g., evolvable hardware, organic computing, and self-monitoring and repairing sensor systems

Readings

Haykin, Neural Networks A Comprehensive Foundation, Prentice Hall, 1998, ISBN 0132733501
G. Cauwenberghs, M. Bayoumi, Learning on Silicon Adaptive VLSI Neural Systems, Kluwer, 1999, ISBN 0-7923-8555-1W.
W. Maas, C. Bishop, Pulsed Neural Networks, MIT Press, 1999, ISBN 0-262-13350-4
D. Mange, M. Tomassini, Bio-Inspired Computing Machines, PPUR, 1998, ISBN 2-88074-371-0
R. Hecht-Nielsen, Neurcomputing, Addison Wesley, 1991

Prerequisites:

Sensor Signal Processing, TESYS

Associated courses :

HEIS

Following courses :

Master project (Diplomarbeit)

Examination:

Oral examination based on semester project. In a computer-based accompanying lab, students become acquainted with employing the presented methods using available Neurocomputers Silimann (analog) and ZISC (digital). FPGA-based platform with VHDL will be considered, too. Topics for semester projects, as individual or group projects will be given to students, which shall be elaborated, documented, and presented. The projects focus on the design and implementation of a real-world task on a Neurocomputer. The work will be presented (20 min. slide presentation, get ISE ppt-style here), discussed, and assessed in the oral examination.

Selected semester projects

Aliye Otan Ebru	Analog Neural Network Hardware for Color Classification	2006
Vural Abdullah	Analog Neural Network Hardware for Color Classification	2006
Stefanie Peters	Weight Optimization for a Neural Network Using Particle Swarm Optimization (PSO)	2006
Max Reichardt	In-the-Loop-Learning Implementierung für analogen Neurochip Silimann 120cx	2006
Jiawei Yang	Digital Neural Network Hardware for Classification	2008
Mahesh Poolakaparambil	Data Classification Using ZISC Digital Neural Network	2008
Ivan Sherbakov	A Parallel kNN Classification Coprocessor	2010

Course number: 85-110

Course materials :

For the course participants, lecture slides (ppt/pdf) and exercises (brief tutorials & tasks) based on available Neuroccomputers Silimann and ZISC as well as Matlab and QuickCog will be provided here. The access information will be proliferated in the lecture. The materials are exclusively made available to the students of this course and can be subject to copyright of different holders without further explicit mentioning. Reprint of the course materials, dissemination of the course materials to third parties, and proliferation of the access information is prohibited without written consent of the course instructor, Prof. A. König, as well as copyright holders of the original materials donated for the purpose of teaching to a limited group.