Noah Fram on Detecting Musical Surprisal
A popular model of brain functions suggests that neurons respond when something doesn't fit their prior expectation---the brain is one big (distributed) change detector. In the past there has been a concerted effort to understand discrepancies using a paradigm known as mismatched negativity (MMN), a roburst neural response to auditory deviations with a characteristic EEG signal. But these responses take a lot of averaging to get a signal. More recent work EEG work has considered the brain in terms of decoding the response in real time. This led to models based on surprisal. Surprisal analysis (from Wikipedia) is an information-theoretical analysis technique that integrates and applies principles of thermodynamics and maximal entropy. And now we can apply these ideas to words and music.
Who: Noah Fram
What: Probabilistic models of neurological responses to violated expectations: The rationale for cross-entropy as a model foundation
When: Friday, June 1 at 10:30AM
Where: CCRMA Seminar Room (Top Floor of the Knoll)
Why: Because our brains like surprises
Noah will be reviewing the literature and proposing a model for music surprisal.
Come to CCRMA and be prepared to be (pleasantly) surprised.
- Malcolm
Probabilistic models of neurological responses to violated expectations: The rationale for cross-entropy as a model foundation
Noah Fram (CCRMA)
Abstract
Expectation-violating stimuli, one of the key tools in the study of perception, can be modelled using information theory as high-entropy parts of a sensory stream that require adjusting an internal model of that stream. These stimuli also consistently elicit pre-attentive EEG and MEG responses, most notably the mismatch negativity (MMN). Recent research has drawn an explicit link between quantitative measures of surprisal in linguistic cues and some of these neurological responses. It seems likely, then, that probabilistic representations of unexpected stimuli could be used to predict the neurological response to a generalized unexpected stimulus. In this talk, I propose the general form of such a model, detail the mechanism for calculating the parameters, propose a validation protocol, and discuss some possible uses of this approach in developing accurate maps from EEG data to real neuroanatomy.
Bio
Noah Fram is a PhD student in computer-based music theory and acoustics at Stanford University’s Center for Computer Research in Music and Acoustics. His main research areas are musical information content, psychoacoustics, and computational musicology. Previously, he studied the perception of musical rhythm with Ian Cross at the University of Cambridge. In addition to research, he is a composer, musician, and lighting designer.