Description: Introduction to machine learning concepts with a focus on relevant ideas from computational neuroscience. Information processing and learning in the nervous system. Neural networks. Supervised and unsupervised learning. Basics of statistical learning theory.
Prerequisites: 311 or consent. Recommended: MATH 307.
Schedule: 15 week syllabus, two lectures per week = 30 Lectures.
Lecture 1: Introductory Lecture: Introduction to machine learning and computational neuroscience.
Lecture 2-3: Basic maths skills: review of important mathematical methods, such as differential equations, linear algebra, calculus of variations, probability theory - specifics depend on student needs.
Lecture 4-5: Basic neuroscience: – Neurons, Synapses. – Information processing and adaptation in the nervous system.
Lecture 6-8: From real neurons to artificial neurons: – Realistic biophysical model: Hodgkin-Huxley model. – Simplifications: ∗ Morris-Leccar model. ∗ Integrate-and-fire model ∗ Threshold units.
Lecture 9-10: Mathematical models of learning: – Hebbian learning. – Spike-timing dependent plasticity.
Lecture 11: From biophysics to machine learning algorithms: the Perceptron algorithm - theory and implementation.
Lecture 12-15: Feed-forward artificial neural networks: – Theory and implementation. – Selected applications to contemporary problems. Subject areas are listed below - specifics are adjusted to student interests.
Lecture 16-20: Support Vector Machines (SVMs): – Introduction to support vector learning. – Selected applications of SVMs. Subject areas are listed below - specifics are adjusted to student interests.
Lecture 21-25: Recurrent Neural networks: – Hopfield network. – Simple recurrent networks. – Selected applications of recurrent networks, including associative memory and content-addressable memory (CAM).
Lecture 26-30: Unsupervised learning: – Undergraduate level introduction to cluster analysis: concepts and different types of clustering algorithms. – K-means and vector quantization. – Applications: image segmentation, data mining, and other areas (see below) - specifics are adjusted to student interests.
Application subject areas include: robotics, computer vision, speech recognition, time series analysis, renewable energies, mathematical finance, geophysics, medical imaging, and bioinformatics. Students are allowed to add subjects.