In March 2020, my research stopped. The cognitive neuroscience lab where I work was shuttered because our research involves collecting behavioral and Techniques for viewing the brain and its activity, especiall... data from human participants – something impermissible by public health directives amid the COVID-19 pandemic. Projects in the lab were paused and despite my initial hope that the stasis would be short-lived, it was looking increasingly like a deep freeze as the months wore on.
Neuroscience is a collaborative discipline. Researchers toil in labs together, students gather in classrooms, and scientists of all career stages gather at conferences to hear about what others have discovered. As the world adjusts to the pandemic, these in-person activities have stopped. But tools that connect people virtually are helping to keep science progressing – online experiments are being conducted and remote conferences being organized. Education is also adapting.
Researchers toil in labs together, students gather in classrooms, and scientists of all career stages gather at conferences to hear about what others have discovered.
This past July I participated in Neuromatch Academy: a three-week, online summer school on computational neuroscience. Neuromatch Academy was a bright silver lining to today’s melancholy state of affairs. What started out seeming like constraints on education imposed by the COVID era ended up being the program’s secret ingredients. The school couldn’t be held in person, but having the classes online vastly expanded the capacity to enroll students and fostered the development of a huge and inclusive scholastic community. While not all sessions could be delivered in real time, having high quality recorded lectures that I could watch at my own pace and that I could return to whenever I wanted helped me to absorb the material. The hands-on lab segments were accessible, only requiring materials that students had in their own homes (i.e., a computer), which complemented the class’ focus on computational methods and mathematical modeling in neuroscience.
To learn more about Neuromatch Academy and how the class came together, I interviewed Dr. Megan Peters, Professor of Cognitive Sciences at the University of California, Irvine, Chair of the Executive Committee, and President of the Neuromatch Academy Organization.
How would you define computational neuroscience to a general audience?
Dr. Peters: Computational neuroscience is trying to understand the brain and its functions using computational approaches. It can range from: “How does this The functional unit of the nervous system, a nerve cell that... talk to that neuron? Let’s write down some equations,” to: “How does this thought process lead to that thought process? Let’s write down some equations”. You’re trying to use computational methods and applied statistics to understand the underlying systems, their causal interactions, and their dimensionality. It’s computational neuroscience as long as there is some level of precise mathematical description.
It can be contrasted with cognitive psychology, which involves a lot of theory and really careful, really beautiful experiments where you’re trying to understand cognitive processes, but not describing them in computational language. Cognitive psychology is a lot of boxes and arrows – this process leads to that process, whereas the computational approach to that would be: What’s inside this box? Let’s write down some equations. And what does this arrow mean? Let’s write down the functional form of that arrow.
What is the role of computational neuroscience in the future of neuroscience?
Dr. Peters: Basic neuroscience is going to become more computational as we go forward. There’s going to be a stronger and stronger mesh between computer science, artificial intelligence, and classic computational modeling – those are going to become a larger and larger part of neuroscience. There’s always going to be room for cellular and molecular descriptions and understanding of how the brain actually functions – how this wetware gives rise to everything we do – but I think it’s going to become more hybridized as we go forward. That’s not to say that’s necessarily only a good thing. I think there’s a lot of obscuring of true understanding that can come with over-reliance on computational approaches, which is why in Neuromatch Academy we tried to focus on: What can this model do for you? How can it answer a specific question, and is that the question you want to answer?
There’s always going to be room for cellular and molecular descriptions and understanding of how the brain actually functions but I think it’s going to become more hybridized as we go forward.
As computational neuroscience becomes a more sophisticated field, there will be a lot of room for applications to real world, clinically relevant outcomes, where we are creating better explanations of why certain biological or chemical changes in the brain lead to this type of change in information processing, which leads to this sort of clinical outcome. I think that is going to be a very tangible benefit, and a very important focus of computational neuroscience going forward. Once you’ve built even a simplified framework – this chemical, or this representation of information, or this configuration of a neural network is why we see this particular clinical or behavioral outcome – then you can go fiddle with your model and use that to drive new clinical interventions, drug discovery, even behavioral interventions to make serious positive impact on the lives of people around us.
How did Neuromatch Academy come about?
Dr. Peters: It started in early April 2020 when Konrad Kording, Gunnar Blohm, Paul Schrater, and I were trying to figure out what to do with CoSMo (Computational Sensory Motor) summer school during COVID. Konrad said: Let’s put it online, invite some other lecturers, and make some videos. We brought in Brad Wyble, who helped Konrad build the Neuromatch conferences. Sean Escola saw the call we put out on Twitter, and came in to help with fundraising. The six of us formed the board of directors for this idea of how to put a computational neuroscience summer school online.
It snowballed from there – we joined forces with organizers of other computational neuroscience summer schools around the world to run one international online summer school instead of multiple fractionated and dispersed schools.
This also presented us an unprecedented opportunity to not have these typically very elite, very competitive to get into, very expensive, in-person summer schools – which can cost $4,500 a student – economically inaccessible, only accessible to the elite students who can get in through the application process because they are already half way through a computational neuroscience or similar degree at an elite university – and even if you can get a fee waiver you cannot pay to fly there, or your lab can’t sponsor you, or even worse – you’re stuck in a country you can’t get out of.
We quickly realized: This is an opportunity to do something that has always been very close to my heart – leveling the educational playing field for everybody. That became a core part of the mission really at the very beginning. This was a real opportunity to change the landscape of what this type of elite education will look like for students worldwide.
This is an opportunity to do something that has always been very close to my heart – leveling the educational playing field for everybody.
Another big piece of the puzzle for us was: A lot of the TAs for these elite schools don’t get paid, which means that you have to be in a very strong position of privilege, to be able to take off for three weeks from your job or child care responsibilities, to be a TA. We paid our TAs a living wage – that was really important to us.
It was a really big production. We had a lecture video editing crew, tutorial editing crew, backend system crew to make it all work together. We had an evaluation team and community management team to monitor student feedback and the students’ experience. We recruited mentors for the students’ group projects. This is like the definition of feature creep. We went from “let’s put some videos up online” to raising several hundred thousand dollars, paying 190 teaching assistants to go through the program with us, and having 1,700 interactive students with an 85% retention rate. I’m still blown away and bowled over by the amount of energy and enthusiasm and love that has been poured into this enterprise. It has grown beyond my wildest dreams.
Do you think Neuromatch Academy can be a model for online education in general, in the era of COVID?
Dr. Peters: We’re writing up a paper right now that’s the first of what should be a long series of papers on what we did, what worked, and what didn’t work so that it can become a model. Somehow being stuffed together for eight hours a day with nine other people in a Zoom box creates camaraderie, community, and a shared intense experience that leads to long-lasting connections, which is also ultimately what you get at an in-person summer school. You spend two or three weeks intensely learning about stuff. Going through that in a group makes you bond in a way that you don’t really get in a lot of other online education. There’s something about the small-group interaction and the intensity that made a big difference there.
Somehow being stuffed together for eight hours a day with nine other people in a Zoom box creates camaraderie, community, and a shared intense experience that leads to long-lasting connections, which is also ultimately what you get at an in-person summer school.
I really hope that this will serve as a model – especially for computationally oriented courses where there’s a hands-on coding component, where it’s not just lectures, but instead it’s: lecture, go do the thing, lecture, go do the thing. Learning by doing and learning by interactive exercise is the right way to go, but learning by interactive exercise isn’t enough on its own. You need to have learning by interactive exercise in a group with cohesion and identity. I hope that will be a lesson for universities going forward and also for other online courses.
Will Neuromatch Academy continue as an online school once public health restrictions are lifted?
Dr. Peters: I don’t see Neuromatch Academy in its current incarnation as a total replacement for in-person summer schools. In-person summer schools serve a very important purpose for early career researchers in terms of building community and making really strong connections that last for years.
What I’d like to see in the future is that this becomes a model for hybrid accessible education, where there’s a standardized curriculum that can be gone through by groups in person together, groups online across regions, and also groups that might not want to do it fully or as intensely. Neuromatch Academy doesn’t have to be only online, but I don’t want to lose the online component either, because that’s what makes it accessible to everybody worldwide – even if you’re stuck behind a geopolitical wall, or you’re stuck behind a fiscal or economic wall.
For students who just completed Neuromatch Academy and want to continue their computational neuroscience journey, what’s a good next step?
Dr. Peters: It depends on which career stage you’re in. You don’t become an expert computational neuroscience researcher in a day. You can learn the tools in say, six weeks, in a hackademy or data science bootcamp. But the way to think about neuroscience and to ask real questions is something that develops not just over the five years of a PhD but over a lifetime.
So if you are coming out of NMA and you are finishing your undergrad now or very soon, then this will provide a very solid foundation that will help you formulate the types of questions that you’re interested in asking during your Master’s or PhD, so you can start to develop intuition about how to ask and answer good questions, and which tools are right for the job.
If you’re a PhD student or Master’s student, see if you can pick some of the techniques that we introduced, use this as the first step. Same if you’re a post doc. Use this as the runway for the next stage. I know that’s kind of a big and lofty, not very concrete suggestion, but I really think that actively working to see how some of these tools and techniques would provide insight into some of the questions that you’re interested in asking will help solidify the techniques, and it will also help you to understand whether the questions you’re asking are the ones you actually want to be asking, and how to maybe change them so they better suit what you’re curious about.
Isn’t computational neuroscience so fascinating? Such methods could open up doors for providing better tools for scientists to analyse the mysteries of the brain.
Interested in more? Give a read to another great article on the rise of artificial neural networks in neuroscience!
Written and Illustrated by Sean Noah. Edited by Nerissa Culi, Alexa Erdogan and Sumana Shrestha.