⭐ Long-term support: Actively maintained for 6+ years, 1700+ stars on GitHub, hundreds of citations (even without a paper!)
🐦 We also release Birdsong NOIZEUS, a new benchmark for bioacoustic denoising

✅ Domain-general: strong baseline when ML models/data aren’t available
🎛️ Stationary & non-stationary variants
⚡ GPU-accelerated for real-time and high-throughput use
🧪 Validated across many domains

Very cool @danielpollak.bsky.social @healeylab.bsky.social - this aligns with what I found in the starling paper we just published using a different context paradigm (sequence integration). NCM (among others) also does not show context modulation but sharpens acuity/attn with expectation

Sensory populations reflect the Bayesian likelihood. And expectation modulates sensory activity. But here’s the twist: sensory neurons don’t integrate the likelihood and prior expectation. (6/n)

Finally, this work is the product of an amazing team, in particular @trevorsupan.bsky.social and Tim Gentner! And a huge thank you to our reviewers and everyone who provided feedback throughout!

This was the longest project of my career! It started nine years ago, back in 2016, the second year of my PHD. Every paper I have ever published has been enveloped by this one.

Takeaway:

1) Song sequence perception follows Bayesian integration.
2) Sensory populations reflect the likelihood, and are modulated by expectation, but don't follow Bayesian integration.
3) Instead, expectation refines sensory precision, leaving an unbiased signal for downstream processing.

This challenges the idea that sensory neurons integrate prior expectations the way decision-making circuits do. Instead, expectation boosts acuity where needed, letting decision-making systems flexibly integrate an unbiased sensory signal. (8/n)

Instead of integrating expectations, neural populations enhance the likelihood of expected stimuli—sharpening perception rather than shifting it.
This means sensory systems maintain a veridical, faithful, representation of the world. (7/n)

Sensory populations reflect the Bayesian likelihood. And expectation modulates sensory activity. But here’s the twist: sensory neurons don’t integrate the likelihood and prior expectation. (6/n)

Behaviorally, birds integrate expectations and sensory input probabilistically, following a Bayesian strategy. This aligns with classic models of categorical perception. 5/n

To investigate this, we trained European starlings to classify ambiguous song syllables generated from a variational autoencoder. We manipulated expectation by changing the probabilities of syllables within song sequences. 4/n

We show that while decision-making systems integrate expectations probabilistically, sensory systems do something surprising: Rather than biasing perception, expectation sharpens it, enhancing sensory precision. 3/n