Figure 1. Model circuitry and brain recordings. Left: Simplified songbird motor control system. Right: Neural activity during singing in the variability generating premotor pathway.

Research context

Songbirds produce learned vocalizations using a well-characterized neural circuit involving the premotor areas HVC and LMAN which both innervate the motor area RA. While the HVC→RA pathway generates highly stereotyped song patterns, input from LMAN onto RA introduces variability that is essential for learning and adaptation. How this variability arises from neural activity — and how it is transformed into structured motor output — remains an open question.

Project description

In this project, we aim to model the song control system by extending a simplified HVC→RA network to include variability-driving input from LMAN. The model will be grounded in experimental data, using neural recordings from LMAN.

A central goal is to investigate how different synaptic nonlinearities shape the transformation from neural activity to vocal behavior. In particular, we will test whether NMDA-like activation functions, such as those proposed by Kim et al. (2023), can account for the observed variability in song.

Tasks

Depending on scope and interest, the project may include:

• Implementing a baseline HVC→RA network model for stable song production

• Integrating recorded LMAN activity as input to the model

• Implementing and comparing different activation functions, including NMDA-based nonlinearities

• Analyzing how variability in neural input affects model output

• Evaluating model behavior against biological observations

Your benefits 

You will gain experience in:

• The song control system of songbirds

• Computational neuroscience and neural circuit modeling

• Working with large-scale neural recordings

• Designing and evaluating neural network models

• Bridging biological mechanisms and machine learning approaches

Your profile 

We are looking for motivated students with an interest in:

• Neuroscience and/or machine learning

• Mathematical modeling and programming

To apply, please send your CV and transcript of records to one of the contacts below.

Literature

[1] Fiete & Seung, 2007, Elsevier

[2] Stark and Perkel, 1999, JNeurosc

[3] Kim et al., 2023, NeurIPS