Who am I?

I’m a neuroscientist by training who is studying how brains learn for my postdoctoral research and is always looking for connections between neuroscience and machine learning. In addition, I love to code, I love sports, and I am a runner, and am working on combining my interests by building and training a deep reinforcement learning agent to play dominoes, the game that I have always played with my family growing up. Other than that, you might find me running through the parks of London, reading voraciously, keeping up to date with American football and endurance sports, and generally trying to learn as much as I can about pretty much everything.

Education and Scientific Training

My journey started a little bit unconventionally. For my undergraduate degree, I studied Saxophone Performance at the Peabody Institute of music. Fortunately, (or by design if you ask my parents), Peabody is a subsidiary of Johns Hopkins University, so when my thoughts turned towards neuroscience, I was able to jump into coursework and research that propelled me to a career in science. I learned how to do research by working with Dr. Charles Limb, where I helped out on an fMRI study of the brilliant musical sauvant Gabriela Montero and wrote a review on the neuroscience of improvisation and creativity.

During my PhD in Bernardo Sabatini’s lab in Boston, MA, I studied how calcium signals evoked by plasticity protocols vary in a dendritic branch-specific manner in cortical layer 2/3 cells of mice using slice physiology, two-photon calcium imaging, and glutamate uncaging. We wrote a paper showing that dendritic morphology shapes action-potential evoked calcium influx by its impact on branch-specific impedance. Afterwards, we explored how these changes to plasticity signals can impact the representational properties of cells, using computational modelling to replicate in vivo data and demonstrate the functional benefits of such a system. We wrote about this in my thesis.

I’m now doing a Postdoctoral Research Fellowship with Kenneth Harris and Matteo Carandini in the Cortex Lab in London, where I’m investigating a biologically-plausible mechanism of backpropagation described here. This project is just getting started, so if you’d like to talk about it, please contact me! I’d love to hear your ideas.

Curiosity-Driven Projects

In addition to my formal training, I’m working on a few other side projects to gratify intellectual curiosities beyond those I satisfy at my job. As a followup to my PhD work, I’ve been thinking about what our biophysical results might mean for neural networks. This brought us down a road towards the network pruning field, where we are trying to come up with new metrics that represent how useful neurons (nodes) are for a network’s function. I’ve been teaching myself deep reinforcement learning by training an RL agent to play dominoes within a custom python environment using pytorch. And I’m building off the dominoes project to do some mechanistic interpretability research on how attention-based networks solve sequencing and planning problems.

Contact

My favorite thing about being a scientist is talking to people. So, really, if you have any ideas you’d like to share or discuss, please contact me! The best way to do so is with my email address- andrewtylerlandau at gmail dot com.