Courses

EE 129. Neural and Nonlinear Information Processing

Description

Catalog Description: (3 units) Three hours of lecture per week. Principles of massively parallel real-time computation, optimization, and information processing via nonlinear dynamics and analog VLSI neural networks, applications selected from image processing, pattern recognition, feature extraction, motion detection, data compression, secure communication, bionic eye, auto waves, and Turing patterns.

Prerequisites: EE120 or consent of instructor

Course objectives: To present a unified treatment of real-time analog computation, image processing, and optimization using analog VLSI neural networks which exploits the massively parallel nonlinear dynamics of the circuit. Novel applications to image processing, waves, and patterns will be emphasized.

Topics Covered:

  • The Hopfield Neural Network
    • Spin Glass and the discrete Hopfield Net
    • Energy function
    • Learning rule
    • Associative memory and storage capacity
    • Analog Hopfield Net and Limitations
  • Cellular Neural Networks
    • Dynamic range and Lyapunov function
    • Linear and nonlinear templates
    • Complete stability criteria
    • Analog to Boolean maps
    • The CNN Perceptron and learning rule
  • CNN as Universal Turing Machine
    • The CNN game of life
    • Architecture of the CNN Universal chip
    • The CNN Compiler and operating system
    • The CNN high-level language
  • Analogic and Morphological Algorithms
    • Erosion and Dilation
    • Edge detection and contrast enhancement
    • Skeletonization
    • Hole detection and filling
    • Connectivity detection
    • Half toning
    • Image enhancement and restoration
    • Color processing
    • Retina and bionic image processing
  • Real-Time Analog Computation
    • Computation intensive mathematical operations
    • Optimization problems
    • Nonlinear programming problems
    • Simulating nonlinear partial differential equations
  • Real-Time Analog Non-Stationary Image Processing
    • Motion detection
    • Motion tracking
    • Velocity estimation
    • Simple speech recognition
  • Nonlinear Dynamics in Electronic Circuits
    • Bifurcation phenomena
    • Poincaré maps and state space dynamics
    • Chaos and strange attractors
    • Lyapunov exponents and time series reconstruction
    • Takens' embedding theorem
    • Auto waves: Concentric, spiral, and scroll waves
    • Turing patterns
  • Applications of Nonlinear Dynamics
    • Pseudo-random number generation on a chip
    • Synchronization and control of chaos
    • Secure communication and cryptography
    • Controlling defects in crystals and lattices
    • High-accuracy optical character recognition
    • Image processing via autowaves
    • Global optimization via autowaves

General Catalog

Undergraduate Student Learning Goals