Sensory coding in strongly correlated neural populations
When
Where
Event language(s)
Welcome to our ABC Seminars! This seminar series is open for everyone. The talk will take place in . After the talks, coffee and pulla will be served.
The event will be also streamed via Zoom at:
Sensory coding in strongly correlated neural populations
Abstract: Sensory neurons are noisy and encode information about input stimuli through their collective activity, which is shaped by two main sources of correlations. First, neurons with overlapping stimulus sensitivity may exhibit signal correlations between their neural responses. Second, network interactions correlate neural noise across the population in a phenomenon termed noise correlation. Strong, positive noise correlations are widely observed across sensory systems, especially among neurons with similar stimulus tuning. Yet, a large body of theoretical work suggests that this combination of both positive noise and stimulus correlations should hinder the encoding of stimulus information. In this talk, I will present a theory of information coding in correlated neural populations that resolves this apparent paradox. We demonstrate, both theoretically and experimentally, that noise correlations enhance sensory coding when they are sufficiently strong. We further show that this benefit arises from enhancing the encoding of fine-grained details of the stimulus, at the expense of large-scale features, which are already well encoded by the network.