Curriculum Vitae

Education

• Feb 2017 - Sep 2021

  PhD, Electrical and Electronic Engineering

  Thesis: Modeling the interaction between the respiratory and the swallowing central pattern generators. Advisors: Prof. Jonathan H. Manton and Dr. Mathias Dutschmann. Awarded the Melbourne Research Scholarship, the Melbourne School of Engineering Studentship, and the Stawell Scholarship.

  The University of Melbourne

  Melbourne, Australia

  ---

• Jan 2018 - Dec 2018

  Deep Reinforcement Learning Nanodegree

  Implemented double dueling deep Q-networks, DDPG, self-playing DDPG, and convolutional classifiers, each solving its benchmark environment to the required score threshold.

  Udacity

  Online

  ---

• Sep 2010 - Jan 2016

  MSc & BSc, Physics — Graduated with Honours

  Specialized in solid state physics of low temperatures and superconductivity.

  M.V. Lomonosov Moscow State University

  Moscow, Russia

  ---

Experiences

• Oct 2021 - Present

  Postdoctoral Research Associate, Tatiana Engel Lab

  Reverse-engineering recurrent network models of cognition and motor control into circuit-level explanations. Showed that the choice of single-unit nonlinearity changes the computational primitives a network builds with, so RNNs trained on identical tasks arrive at structurally distinct circuits that diverge out of distribution (Nature Machine Intelligence, 2025). Proved that in ReLU networks the selection vector’s support can only change through suppression of units receiving irrelevant stimuli, yielding a connectivity rule that predicted targeted rank-one perturbations of trained networks. Author of trainRNNbrain, an open-source package for training and dissecting task-optimized RNNs.

  Princeton Neuroscience Institute

  Princeton, NJ (Cold Spring Harbor Laboratory until Jan 2023)

  ---

• Sep 2023 - May 2024

  Shenoy Undergraduate Research Fellowship in Neuroscience (SURFiN) Mentor

  Mentored an undergraduate fellow through a full research cycle, from problem formulation (reverse-engineering the circuitry behind a Go/NoGo task) to a conference abstract and a poster presented at the SURFiN conference.

  Simons Foundation

  New York, NY

  ---

• Mar 2017 - Nov 2020

  Teaching Assistant and Tutor

  Ran workshops for Signals and Systems, Probability and Random Models, and Signal Processing, teaching more than 50 master’s students per semester. Tutored undergraduates in Real Analysis, Calculus II, Quantum and Thermal Physics, and Probability Theory at Janet Clarke Hall and Newman College.

  The University of Melbourne

  Melbourne, Australia

  ---

Last publications

• Modelling of synaptic interactions between two brainstem half-centre oscillators that coordinate breathing and swallowing

  Authors: P. Tolmachev, R. R. Dhingra, J. H. Manton, M. Dutschmann

  Biological Cybernetics (accepted, to appear) • 2026

  A model of the synaptic interactions between the respiratory and swallowing half-centre oscillators in the brainstem, producing testable predictions for breathing–swallowing coordination.

• Single-unit activations confer inductive biases for emergent circuit solutions to cognitive tasks

  Authors: P. Tolmachev, T. A. Engel

  Nature Machine Intelligence, 7, 1742–1754 • 2025

  The choice of single-unit nonlinearity determines which computational primitives a network builds with: RNNs trained on identical tasks and matched on in-distribution performance arrive at structurally distinct circuits and diverge behaviorally under previously unseen stimuli.

• Linking neural dynamics and circuits: selection vector re-orientation via node inhibition

  Authors: P. Tolmachev, C. Langdon, T. A. Engel

  In preparation — contributed talk, COSYNE 2023 • 2023

  In ReLU networks the selection vector’s support can change only through suppression of units receiving irrelevant stimuli — a concrete circuit implementation of a mechanism previously described only at the population level. The resulting connectivity rule predicted rank-one perturbations that realigned selection vectors and selectively degraded behavior in one context while sparing the other. Equal contribution with C. Langdon.

• New Insights on Learning Rules for Hopfield Networks: Memory and Objective Function Minimisation

  Authors: P. Tolmachev, J. H. Manton

  International Joint Conference on Neural Networks (IJCNN), 1–8 — talk at WCCI 2020 • 2020

  A unified normative perspective on local learning rules for pattern memorization in Hopfield networks, yielding new rules that raise pattern-storage capacity over the standard Hebbian rule.

• A Distributed Transmission Scheduling Algorithm for Wireless Networks Based on the Ising Model

  Authors: X. Li, P. Tolmachev, M. Pauley, J. H. Manton

  IEEE Statistical Signal Processing Workshop (SSP), 6–10 • 2018

  A distributed scheduling algorithm for wireless transmission derived from the Ising model.

• Modeling the Respiratory Central Pattern Generator with Resonate-and-Fire Izhikevich Neurons

  Authors: P. Tolmachev, R. R. Dhingra, M. Pauley, M. Dutschmann, J. H. Manton

  International Conference on Neural Information Processing (ICONIP), Lecture Notes in Computer Science, 603–615 • 2018

  A spiking model of the respiratory central pattern generator built from simplified Izhikevich neurons, reproducing the three-phase breathing rhythm with greater clarity and at lower computational cost.

Skills