- Course Number: 16:198:525
- Course Type: Graduate
- Semester 1: Fall
- Credits: 3
The course explores how computation in the human brain can be effectively modeled across the main levels of abstraction (from a single neuron to biological neural networks and system; introduces a time-sensitive computational formalization of brain function based on the model of neuron as a Spike Processing Machine – Spike Neural Networks (SNNs); and employs neuro-mimetic or neuro-inspired SNNs to tackle a problem in a term-wide project. The aim of this course is to provide the student with a solid foundation in the field.
- M.S. Course Category: AI/Machine Learning
- Category: B (M.S.), B (Ph.D.)
- Prerequisite Information:
01:198:205 or 01:198:440
- Course Links: 01:198:205 - Introduction to Discrete Structures I, 01:198:440 - Introduction to Artificial Intelligence
* Elements of Neuronal Dynamics in Biophysically realistic Neuron Models* Dimensionality reduction and phase plane analysis: Integrate & Fire, Izhikevich, Spike Response Models, Nonlinear Neuron Models* Parameter Optimization in linear and non-linear models* Evolving Neuronal Populations* The Brain as an optimization machine* Learning via synaptic tuning* Memory and Attractor Dynamics* Synaptic Plasticity and Learning in Spiking Neural Networks* Computational elements of decision making, emotions and consciousness
Suggested Textbooks: (1) Keith L. Downing “Intelligence Emerging: Adaptivity and Search in Evolving Neural Systems” MIT Press | May 2015 (2) Peter Sterling and Simon Laughlin “Principles of Neural Design” MIT Press | March 2015 (3) Dana H. Ballard “Brain Computation as Hierarchical Abstraction” MIT Press | February 2015
- Expected Work: A term project on Spiking Neural Networks; 2 assignments - to prepare the students for the term project; Paper presentation.
- Exams: The usual practice is a term-project.
- Learning Goals:
The course provides an overview of the fundamental concepts and current trends in Neuro-morphic Computing with a focus on designing spiking neural networks for robotic vision and movement.