Master Project: A Neural Kalman Filter
We (Grewe and v.d. Malsburg) are looking for a student (Master level) who is interested in carrying out a theoretical neuroscience project at the Institute of Neuroinformatics. Consistency between sensory input and model-based predictions of our brain is the basis for the confidence with which we rely on our senses. Achieving this consistency is also goal and basis of learning. Sensory noise and uncertainty in latent model parameters make it important to deal with probability distributions. In the technical domain this is handled by Kalman filters, see Figure 1. In these, probability distributions are parameterized by Gaussians. In this project we would like to work out a novel idea of achieving Kalman filter-like functionality based on activity distributions over rows or fields of neural value units and a set of differential equations for the dynamic evolution of these distributions. The task of the master project is to apply this idea to a problem of visual motion detection.
• Literature search
• Development of a system simulation program
• Application to structure-from-motion problem
• Performance of tests
What we offer
• The interdisciplinary and collaborative environment at the intersection of engineering, neuroscience, and artificial intelligence
• A highly motivated research team and cutting-edge research project
• The potential for continuing work at the INI
Start of Project: Immediately
Length of Project: 6-9 months, Contact: bgrewe (at) ethz.ch and chvonde (at) ini.uzh.ch
Interested students with a background in neuroscience, computer science, physics, mathematics or similar are encouraged to contact us. Please attach a CV, short motivation and background (<0.5 page). If you have any questions about the project, do not hesitate to contact us.