neuralreckoning

neuralreckoning, 1 year ago to random

To UK academics (students and staff).

⚠ If you're offered work marking in the next months, please say no!

It's likely an attempt to break our marking and assessment boycott, our most effective industrial action to avoid threatened pay cuts.

Please boost if you have UK followers.

#ucu #ucurising #academicchatter

neuralreckoning, 7 days ago to random

OK, let's try something new. I'm not well connected because I'm bad at in person networking, and this is compounded by my decision to stop flying to conferences. So, can I use mastodon to find potential experimental colleagues who would like to work together?

Ideally for me, this would be people in Europe so I can visit by train, but it's not essential. I have some ideas for interesting projects and grant applications, and I'd love to develop those into concrete projects in close participation with experimental colleagues.

One of the main themes I'm interested in is how we can relate various neural mechanisms (e.g. inhibition, recurrence, nonlinear responses) to functions, using computational modelling to ask 'what if' questions that couldn't be answered by experiments alone.

I'm also interested in thinking about how we can use "information bottleneck" ideas to think more clearly about what computations networks of neurons are doing, going the next step beyond representing information to computing / discarding information.

A big question I'd like to answer is to find out how different brain regions work together in such a flexible and scalable way.

A technique I'm very excited about at the moment is using modern ML algorithms to train spiking neural networks at cognitively challenging tasks, making them directly comparable to both psychophysical and electrophysiological data.

Part of that could involve building in new mechanisms, like dendritic structure or neuromodulators into those networks and allowing the trained networks to make use of them in the best way possible.

I'd also love to build jointly motivated experimental and theoretical/synthetic datasets to test models against.

If any of that sounds interesting to you, take a look at some of my recent papers and get in touch. I'd love to hear from you.

http://neural-reckoning.org/publications.html

neuralreckoning, 2 months ago to random

"If you want to minimize the possibility of unexpected breakthroughs, tell [scientists] they will receive no resources at all unless they spend the bulk of their time competing against each other to convince you they already know what they are going to discover." - David Graeber, The Utopia of Rules.

neuralreckoning, 2 months ago to random

Asimov's first Foundation novel has a wonderful scene that I think prefigures the LLM arms race we're going through at the moment. A bunch of characters are using formal mathematical tools to analyse the meaning of verbose and seemingly eloquent political statements:

"That," replied Hardin, "is the interesting thing. The analysis was the most difficult of the three by all odds. When Hoik, after two days of steady work, succeeded in eliminating meaningless statements, vague gibberish, useless qualifications - in short, all the goo and dribble - he found he had nothing left. Everything canceled out."

neuralreckoning, 4 months ago to Neuroscience

We made a free, open course on #Neuroscience for people with a #MachineLearning / quantitative background. Get deep in #ComputationalNeuroscience or just indulge your neurocuriousity. I believe that neuro and ML can learn from each other and do better together than on their own.

The course has 34 short videos from introductory topics right up to recent discoveries we still don't fully understand. We also have practical exercises focussed on open ended discovery, fully compatible with Google Colab.

Check out the course website at:
https://neuro4ml.github.io/

My thanks to brilliant co-developer @marcusghosh, and contributors @GabrielBena, Swathi Anil and Greta Horvathova.

Over the next year, I'll be turning this into an 'interactive textbook' with videos, text and runnable code in one place, and welcoming contributions on new topics, corrections, etc. through GitHub issues. All our materials are freely licensed for reuse in your own courses too.

Why this new course? There's a lot of intro neuroscience courses out there, and a lot of ML for neuroscientists, but I wanted this one to be specifically for quantitative people who are curious about the brain, how it might be similar and different to ML.

I hope you'll enjoy it!

neuralreckoning, 7 months ago to random

This week I read about a Nobel winner whose groundbreaking work didn't get funded and got her demoted, and about data fraud by two of the highest profile scientists who were lauded and mega funded. We have to stop rewarding short term flashy work and overproductive scientists.

It's fine and correct to talk about both incentives and individual responsibility. But if we scientists collectively decided to heavily downplay work without open, raw data and reproducible methods, and ignored journal title when evaluating scientists, this couldn't happen.

The system is absolutely broken and needs structural reform, yes. Journals need to go. Competitive grants are the wrong way to fund science. Scientific prizes are very problematic. But we also need to get better at reading and doing science and valuing what works in the long term.

That's the key point. If we let these things happen it means we are doing science badly.

neuralreckoning, 6 months ago to random

I can see why gpt might seem good for coding and writing but I don't want to write code or text that I don't understand. Anything it generated I'd have to go over really carefully to make sure I understood or agreed with. I don't see this as likely to take less time than just writing it in the first place. The bottleneck for me is not the typing it's the sense making and I don't trust gpt with that.

neuralreckoning, 2 months ago to random

How can we fix academic publishing? I just wrote a new article outlining my thoughts on this based on all the attempts I've seen, what has worked and what has failed, and finishing with the strategy we developed for @ScholarNexus. I'd love to hear your feedback!

https://thesamovar.github.io/zavarka/how-do-we-fix-publishing/

neuralreckoning, 3 months ago to Neuroscience

I'm on the latest episode of Brain Inspired talking about #Neuroscience, #SpikingNeuralNetworks, #MachineLearning and #Metascience! Thanks Paul Middlebrooks (not on Mastodon I think) for the invite and the extremely fun conversation. For the explanation of why this picture you'll have to listen to the episode. 😉

https://braininspired.co/podcast/183/

Also, if you're not yet listening to Brain Inspired you should be - and support Paul on Patreon. He provides this free for the community with no adverts. What a hero!

neuralreckoning, 10 months ago to Neuroscience

New (updated) #preprint!

Defining #neural #modularity is hard: much history. We used toy ANNs to show structural and functional definitions not tightly related, resource constraints important, and we need to start thinking about temporal dynamics.

🧵 with @GabrielBena #neuroscience

https://arxiv.org/abs/2106.02626

neuralreckoning, 9 months ago to random

If you're at #CCN2023 and interested in multimodal processing, come and chat with us at our poster P-3B.68 with @marcusghosh from 1-3pm today in the Marquee.

https://2023.ccneuro.org/view_paper.php?PaperNum=1062

If you're not at CCN, check our thread/preprint and we'd be happy to discuss!

https://neuromatch.social/@neuralreckoning/110785811144218374

neuralreckoning, 7 months ago to Neuroscience

I'm happy to announce the start of a new free and open online course on neuroscience for people with a machine learning or similar background, co-developed by @marcusghosh. YouTube videos and Jupyter-based exercises will be released weekly. There is a Discord for discussions.

For more details about the structure of the course, and to watch the first video "Why neuroscience?" go straight to the course website:

https://neuro4ml.github.io

Currently available are videos for "week 0" and exercises for "week 1", but more coming soon.

Why did I create this course? Well, I think both neuroscience and ML can be enriched by knowing about each other and my feeling is that a general purpose intro to neuro or comp-neuro isn't the right way to inspire people in ML to be interested in neuro.

I hear a lot about neuroscience inspiring AI, but I think there's understandable scepticism about that from ML people. I don't want people to take neuro ideas and apply directly to ML, I just think we get a richer picture of what both fields are doing if we think more widely.

In other words, we should be thinking that we are somehow studying the same problem in different ways. You see that in the early history of the field, and it's very inspiring. (Yes, this is pretty much just saying that cognitive science is cool, but my scope is a bit narrower.)

The focus then is not on how neuroscientists think the brain works, but on the mechanisms the brain uses. These are strange, inspiring, and often their contribution to intelligent behaviour is still deeply mysterious.

The first video of the main part, on the structure of neurons, finishes with recent research (from @ilennaj and @kordinglab among others) on what the function of dendritic structure might be. No answers, just ideas.

And that's going to be another key part of this course. Research level problems are not hard to find in neuroscience, and the aim of this course is to empower students with the tools to start finding and working on them straight away.

Most of the exercises in the course won't have correct answers. They're starting points for further investigation. We'll be downloading and exploring open neuroscience datasets using methods from computational neuroscience and ML.

The course is not supposed to be comprehensive. It's a short course and the aim is more to get inspired and start on a longer road. I'd expect everyone to get something different out of it, and I'm happy if for some people their take home is "neuroscience is not for me"!

In some ways, it's the course I would have liked to get me into neuroscience and for my incoming PhD students from non-neuro backgrounds to be able to take. It's personal, and full of the sort of stuff that inspires me to be interested in neuroscience.

Well, I hope that some of you might be interested to follow along in the next few weeks, and since it's the first time I'm giving this course please do give feedback by email, Discord or however you like. Also, please feel free to re-use materials however you like.

#neuroscience #compneuro #machinelearning #ML

neuralreckoning, 1 month ago to random

Early prototype of Mastodon thread viewer:

https://thesamovar.github.io/masto-thread-view/test.html

Just paste the URL of the thread into the box at the top and hit the "linear thread view" button below and it will give you a view of the thread with hierarchical replies sorted by how many engagements they got (reposts + favourites + replies).

It's very early days so it doesn't yet show any images, the design is not ideal, not optimised for mobile, etc. But I already find this useful for getting a feel of big threads.

My aim here is to give people a better way to navigate overwhelmingly large threads and to allow for a sort of archive of interesting threads. If we want to make Mastodon into a viable option for having scientific debates (e.g. alternative to peer review), we need some way to make them more accessible to outsiders and to surface the most interesting and relevant content.

So I'm particularly interested in hearing suggestions for features or other ideas on how to display threads in the context of long lasting discussions with some permanence to them.

At the moment it's just a very simple idea, but I have other ideas for how to display threads that are a bit wackier and I'll add these as extra buttons as and when I work on this. I'm also going to see how feasible it is to make this into a bookmarklet so you can just hit the 'render thread' bookmark in your browser and open a tab with this. Should be straightforward.

If you're interested, please feel free to post suggestions and issues either here or on github: https://github.com/thesamovar/masto-thread-view

May be of interest to @NicoleCRust @jonny

neuralreckoning, 6 months ago to random

Puzzled: We learn the most when we take our understanding of how something works, turn it into a concrete model, make predictions and then those predictions are shown to be wrong experimentally. But this wouldn't get published. We prefer new results predicted well by new model.

Are we maximising model quality at the cost of slowing down the rate of gaining new understanding?

neuralreckoning, 3 months ago to random

I've been thinking a lot about how we could have a non-hierarchical science, and one idea has crystallised.

The way science is done now, senior scientists have a lot of decision making power: which papers get published, which grants get funded, who gets hired. This introduces a hierarchy and concentration of power that has both social problems (bias, well documented potential for abuse of trainees), as well as scientific ones (ideas that challenge old ways of thinking have a much harder time than they should).

However, I wouldn't want to entirely eliminate the collective expertise of senior scientists. It's always amazed me just how well some of them can cut through nonsense and see to the heart of an issue. I distinctly remember enthusiastically going to one of my postdoctoral advisors to talk about my latest complicated modelling idea and getting the response "yeah you could do that but what would it tell us about X?" and realising that they were completely right. I avoided months of fruitless work thanks to that one ten minute conversation.

But do they need to have decision making power to do that? I don't think so. We should give decision making power to junior scientists: they should decide what ideas they work on, how to carry out their research, where to do it, who to collaborate with, and what to publish. The additional role of senior scientists is to give the junior scientists advice, which those junior scientists are entirely free to ignore. You don't need to force people to listen to advice. If the advice is good, freely given and not binding, people will seek it out. And there's no reason it has to only be senior scientists who are in this advice giving role, and no reason that as a senior scientist you need to be in this role if you don't want to be.

This inverts the power dynamics in a really progressive way. With this approach, there's no way to impose your idea of how science should be done on anyone, instead you have to persuade them. This is exactly how it should be. By placing arbitrary authority at the heart of science we've made it unnecessary for established ideas to argue for their value, because the holders of those ideas can just deny publication, grants and jobs to those who disagree. Why bother arguing when you can do that?

An obvious follow-up question is: OK, but then how do you allocate funding? It's a good question and one I'm happy to discuss ideas about. But it's not a case of us having a good answer already and needing a strong argument for an even better way. The current system is a hierarchy whose very nature is contrary to the basic values of science. I suspect almost any alternative would be better. Personally, without a clear winner in mind, I suspect the best approach would be heterogeneous: let's try out different ideas and see what works instead of all the countries in the world converging more or less on variations of this same basic formula.

neuralreckoning, 10 months ago to random

"It is part of the ERC’s core business to carefully select the best and most creative researchers from all fields in its grant competitions."

That started me wondering: is there any evidence that this is a valid approach?

For example, is there any evidence that giving funding to PIs to choose which postdocs to hire is more efficient than directly giving funding to postdocs and letting them choose who they want to be supervised by?

I'm glad that the ERC is making some small steps towards improving the way they evaluate research but I wonder if it's not missing a much bigger picture here. I've never seen any evidence one way or the other on this, but would be happy to be shown any if there is some.

https://erc.europa.eu/news-events/magazine-article/research-assessment-ScC-view

neuralreckoning, 7 months ago to random

Here's me talking about multimodal processing and spiking neural networks at the VVTNS seminar series - many thanks to the organisers for the invite!

https://youtu.be/jnQpnyASJe8?si=aR9QTgDYzs8A2xqG

neuralreckoning, 2 months ago to random

Can anyone explain something about university finances to me? The finance people say that grant overheads don't actually cover the cost of research. I would take this to mean that each grant is a net financial loss to the university. So why do they want us to get more grants?

neuralreckoning, 7 months ago to random

Idea for a new type of online "conference". Instead of focussing on talks, the idea would be to collectively come up with an annual, short, readable review of new work and trends in the field over the last year. Could that work? Would you be interested in taking part if it could?

neuralreckoning, 1 month ago to random

If you're considering a life in academia it's worth watching this video and deciding if it's worth it to you or not. All of this is true.

https://www.youtube.com/watch?v=LKiBlGDfRU8

For me the answer is yes, despite all the problems, for two reasons.

Firstly, I'm lucky enough that I do have considerable freedom to work on the things that I'm interested in. If I was more interested in success or if I was on a 'soft money' position and forced to chase constant grants, I don't know if that would be true. But, such luck is rare.

Secondly, as a socialist I would feel very uncomfortable spending my creative energy on most of the non-academic things I'm qualified for: advertising and surveillance (i.e. tech companies), finance, or startups (making venture capitalists even richer). I could imagine academia getting bad enough that I'd make that choice, but for me it's not there yet. I completely understand that it is that bad for others and I mean no criticism of them.

In a way I suppose this is a sort of defence of academia, but it's a half hearted one at best. I think it's absolutely tragic and depressing that academia has become like this. Doing research should be one of the most joyful and creative things anyone could do with their lives.

neuralreckoning, 1 month ago to random

My university will hire a couple of undergrads to help turn my neuroscience course https://neuro4ml.github.io/ into an interactive textbook along the lines of @neuromatch academy https://compneuro.neuromatch.io. I'd like to try a little more though.

I'd like to write some extensions to JupyterBook so that I don't have to maintain separate slides but have everything integrated into one structure, including video recordings, so that you can be watching like a lecture, pause and you're already at the code you can run, etc.

Anyone seen anything like this done before? Any tips? Any thoughts on how to do it? Features it should have? Be ambitious! In my head I'm calling this "textbook of the future" just to give you an idea of how grandiose you should be. 😉 Cc @choldgraf @rowancockett

neuralreckoning, 3 months ago to random

"the challenges that science is experiencing now ... are due to a lack of emphasis on ... the hard intellectual labor of choosing, from the mass of research, those discoveries that deserve publication in a top journal"

🤔

https://www.science.org/doi/10.1126/science.ado3040

neuralreckoning, 3 months ago to random

So parents of Mastodon, we all do that thing where we're so tired and sleep deprived that we fall asleep while reading a story to one of our children, only to wake up and realise that even though you were asleep your brain was still reading out loud? And that because you were asleep you weren't reading the actual story but just improvising some complete gibberish that the child was totally happy with? It's not just me, right?

neuralreckoning, 1 month ago to random

Fellowship opportunity for ECRs <4y post-PhD. Engineering. 5 year funding. Significant advantage given to people from "underrepresented groups" (see below). Internal deadline of May 6. Comp neuro has done well recently in our dept. Email me if interested.

https://raeng.org.uk/research-fellowships

"The Academy has identified the following groups that are currently clearly significantly underrepresented in UK engineering research:
• Women
• Black people, including those with any mixed ethnicity with Black ethnic background(s)
• Disabled people"

The application process is that candidates need to apply initially to our department, who will select up to 2 applicants to submit to the university as a whole, who select up to 4 applicants (at least 2 from underrepresented groups) for the national competition.

This is an engineering post so it can't be pure neuroscience, but we've had good success recently getting engineering fellowships for @marcusghosh (multimodal processing in the brain with possible applications) and @danakarca (spatially distributed spiking neural networks).

neuralreckoning, 1 month ago to random

Thread viewer for Mastodon. If you're looking for an easier way to navigate deep threads with many contributions, check out "mastodon thread viewer" (early beta for the moment).

https://thesamovar.github.io/masto-thread-view/

The way it works is you bookmark one of the view types on that page, and then when you're viewing a post from a mastodon thread in your browser, simply click your bookmark and it will open a new tab with the page you're currently viewing rendered as a thread (either tree or table view).

It's early days so there may be bugs, etc., but I think it's already useful. Please give feedback on bugs/feature requests either here or via issues at https://github.com/thesamovar/masto-thread-view.