When I tell people I study “computational neuroscience,” I usually get one of two reactions: a puzzled look followed by “so… you study computers?” or an excited “oh, like brain-computer interfaces?” Both responses miss the mark, but they’re understandable — it’s a genuinely strange and wonderful field that sits at the intersection of biology, physics, mathematics, and computer science.
This post is my attempt to explain what computational neuroscience actually is, why I fell in love with it, and what the first years of a PhD in this field have taught me.
What is Computational Neuroscience?
At its core, computational neuroscience asks: what are the computational principles that allow the brain to process information?
We don’t just want to catalog the brain’s anatomy or chemistry (though that matters enormously). We want to understand the algorithm — the logic that explains why the brain is built the way it is and why it produces the behaviors it does.
This leads us to build mathematical models of neural systems. These models range from extremely biophysically detailed simulations of individual neurons (accounting for ion channels, membrane potentials, dendritic computations) to high-level abstract models (like rate-coded networks or probabilistic inference frameworks).
The goal is always the same: find the minimal set of assumptions that explains the data and generates testable predictions.
Why I Was Drawn to It
I came from a mathematics background and was always captivated by the idea that nature has “deep structure” — that the messiness of the biological world is underpinned by elegant mathematical principles. When I discovered neuroscience, I realized the brain might be the most compelling example of this.
Consider these facts:
- The visual cortex encodes orientation, spatial frequency, and color using population codes that can be described geometrically
- The hippocampus appears to implement something like manifold learning to create cognitive maps of space and experience
- Decision-making circuits look suspiciously like Bayesian inference machines
- Learning in the brain shares surprising similarities with gradient descent in artificial neural networks
Each of these observations suggests that the brain is doing something deeply mathematical. I wanted to be part of the effort to figure out exactly what.
What the First Years of a PhD Have Taught Me
1. The Data Are Messy, and That’s the Point
I came in expecting clean, beautiful datasets where patterns would jump out immediately. Instead, I found noisy spike trains, trial-to-trial variability, and recordings that drift over time. Learning to embrace noise — to model it rather than eliminate it — was one of my first and most important lessons.
Noise in neural data isn’t just a measurement artifact. It reflects fundamental properties of neural computation, like stochastic synaptic transmission and the brain’s strategies for dealing with an inherently uncertain world.
2. You Need Both Depth and Breadth
Computational neuroscience requires you to be fluent in multiple languages: the language of neurobiology, the language of mathematics, and the language of computation. You don’t need to be a world expert in all three, but you need to be able to communicate across all three.
I’ve had to read papers from information theory, dynamical systems, Bayesian statistics, and cell biology in the same week. It’s overwhelming at first, but over time you start to see the connections.
3. The Best Questions Come From Experiments
There’s a temptation, especially coming from a math background, to start with a beautiful mathematical framework and ask “what can this model explain?” But the most exciting and important questions almost always come from puzzling experimental observations.
Some of my favorite research directions emerged from asking: “Wait, why does the data look like that?”
4. Collaboration Is Everything
Computational neuroscience is inherently collaborative. Theorists need experimentalists to ground their models in reality. Experimentalists need theorists to make sense of complex datasets. The field advances fastest when these two communities talk to each other.
I’ve been incredibly fortunate to work with lab mates who span this spectrum, and the cross-pollination of ideas has made my research immeasurably richer.
What I’m Working on Now
Right now, my research focuses on understanding how neural populations in sensory cortex represent uncertainty. When the brain receives ambiguous input, it doesn’t just produce a single “best guess” — it seems to maintain a distribution of possibilities. How is this distribution encoded in neural activity?
This question connects to larger themes in Bayesian brain theories and has practical implications for how we think about perception, decision-making, and even certain psychiatric conditions.
It’s a hard problem. Most days I feel more confused than when I started. But that’s also the sign of a good problem.
Looking Forward
If you’re curious about the brain, and especially if you have a quantitative background, I can’t recommend computational neuroscience highly enough. It’s a field where your intuitions are constantly challenged, where the problems are genuinely hard and genuinely important, and where you’re always surrounded by brilliant, curious people.
I’ll keep sharing my journey here — the papers I find exciting, the techniques I’m learning, and the questions I haven’t figured out how to answer yet.
Stay curious. 🧠