Class Unlearning via Depth-Aware Removal of Forget-Specific Directions Machine unlearning aims to remove targeted knowledge from a trained model without the cost of retraining from scratch. In class unlearning, however, reducing accuracy on forget classes does not necess...

Congratulations to my great students Arman Hatami and Romina Aalishah on their new study of machine unlearning and forgetting accepted.

arxiv.org/abs/2604.15166

This has had clear implications for ML and neuroscience (albeit from different angles), and we continue this work in both disciplines….

I am deeply grateful to the Bloomberg Endowment for supporting both our research and broader vision! We hope that our work will contribute to next generation of psychiatry treatments, bci, and ai development.

Great Bita

Temporal-orbitofrontal pathway regulates choices across physical reward and visual novelty Novelty can directly impact valuation of future outcomes, affecting how we choose among rewards such as food or money. This is the case even when novelty is objectively valueless. Here, Ogasawara et a...

! New paper from the lab !

How does novelty impact value-based decision making ?

What are the circuits and what mechanisms do they implement?

Congratulations to Dr. Takaya Ogasawara and team

www.cell.com/neuron/fullt...

Integrator dynamics in the cortico-basal ganglia loop for flexible motor timing - Nature How the mouse frontal cortex and striatum interact to implement integrator dynamics controlling movement timing is explored.

Nature research paper: Integrator dynamics in the cortico-basal ganglia loop for flexible motor timing

go.nature.com/4oX2rmq

incredible collaborating with you Naoki.

hanging out in your office and discussing the details was really fantastic and i gained much insight through it.

Ventral striatal astrocytes contribute to reinforcement learning Astrocytes influence synaptic plasticity and neuronal function through astrocytic calcium dynamics (ACD). However, astrocytes contribution to cognitive operations like reinforcement learning (RL) rema...

How do non-neuronal CNS cells called astrocytes contribute to cognition & computation?

See our preprint

Headed by my students Julia Pai and Fatih Sogukpinar, also working with Hiratani, Pignatelli, Papouin, Frank, and Ching labs. This was a huge effort

www.biorxiv.org/content/10.1...

she is brilliant

Excited to work with you and the Dept! Challenging times but science remains truly thrilling!

thanks for your kindness, support, and friendship

At JHU, our team will work on the circuit / algorithmic basis of intelligence & cognition, with attention to their malfunction in disease. We are looking for new team members with experience in circuit neuroscience, biology, or machine intelligence. Please DM if u are interested.

Some professional news : My team is moving to Johns Hopkins University this Summer. I am deeply grateful to Washington University for supporting us since 2014. It has been an amazing journey full of brilliant colleagues and friends.

Absolutely. But not “disorganized” either. Very particular pathways with likely somewhat dissociable learning rules (eg., timescales) supported by distinct cell types and modulators - that’s roughly my bet

Temporal-orbitofrontal pathway regulates choices across physical reward and visual novelty Perceptually novel objects have profound impacts on our daily decisions. People often pay to try novel meals over familiar ones, or to see novel visual scenes at art exhibits and travel destinations. ...

New work!

How does perceptual novelty impact the value of (physical) rewards: what are the underlying algorithms and neural mechanisms?

The answers may help to construct smarter AIs and to understand how and why novelty seeking breaks in psychiatric disorders.

www.biorxiv.org/content/10.1...

Psilocybin’s lasting action requires pyramidal cell types and 5-HT2A receptors - Nature A pyramidal cell type and the 5-HT2A receptor in the medial frontal cortex have essential roles in psilocybin’s long-term drug action.

congrats @alexkwan.bsky.social

www.nature.com/articles/s41...

First blue sky post is to celebrate! Incredibly proud of Dr Julia Pai. Fantastic defense and pioneering project on non-neuronal cells in Reinforcement Learning! Now almost off to her next chapter. Paper in prep!