Ask HN: Designing your own PhD?
Hey HN
I'm just wrapping up a double masters in neuroscience and machine learning, and I was wondering if it was possibly to basically do the equivalent work of a phd, publish 5-6 papers on your own?
The topic I'm really getting interested in is at the intersection of Wetware Computing (https://en.m.wikipedia.org/wiki/Wetware_computer), Neuromorphic Computing, and Deep Learning. I've found some programs like the iPhd at University of Washington, but I can't find too many professors who would be willing to allow me to work on these topics on my own.
Would it be feasible to work on my own for a few years, assuming finances weren't an issue, and publish a few noteworthy papers and open source work? Would that basically be the equivalent of what a PhD is supposed to signal?
I'm a professor in pure mathematics (at a university with a Ph.D. program), in the United States. I'm a little bit biased, since I went through the system, but that said --
> Would that basically be the equivalent of what a PhD is supposed to signal?
Signal to whom? Probably not, in my estimation, but maybe -- it depends on what you're attempting to accomplish.
Enrolling in a formal, traditional Ph.D. program has lots of benefits; getting to work under a mentor is only one of them. Perhaps the most significant is that you become part of a network of peers: you surround yourself with people whom you can bounce ideas off of. Also, there will generally be lots of visitors passing through, with lectures and opportunities to go out to dinner with the speakers afterwards -- also invaluable learning perspectives.
Unfortunately interdiscplinarity is a difficult thing: most faculty only know their own discipline. (University administrators love interdisciplinary research, but impressing them is irrelevant before you have a tenure-track position.)
> and publish a few noteworthy papers
Noteworthy to whom?
It sounds like you want to pioneer a new research area that nobody works on. Would you be satisfied to do work that, in some abstract sense, is worthy of taking note of? Or is it important to you that other people actually notice? If the latter, then you almost certainly want the backing of an advisor and of an institution.
In general, professors tend to appreciate students who take the initiative and seek out their own research projects. Depending on your long-term goals, I would recommend seeking out professors who work in one of the topics you mentioned, and depending on their response, consider enrolling in their Ph.D. programs.
Best of luck!
CS Ph.D. here. Yes, it's possible, but I would strongly recommend against it. The main reason is that you are almost certainly not as smart as you think you are, and so trying to advance the state of the art in human knowledge (which is what a Ph.D. is supposed to be all about) on your own is very likely to fail. This would be true even in a perfect world, and the world we actually live in is far from perfect. The process of publishing is fraught with politics, and being affiliated with an institution is a not-insigificant advantage in and of itself, completely apart from the benefits of actually being on the inside and having access to people to whom you might not otherwise have access to. Trying to do a Ph.D. equivalent on your own is like trying to climb Everest on your own with a 50-lb dead weight attached to your ankles. It might be possible, but the odds are very much against you.
I would advise you to search around a bit more thoroughly for universities who are doing such work.
I don't know much about CS PhD programs, but in biomed, solo publications are very rare, just due to the amount of resources needed. Also, you lose the credibility of an institution, and your work will be more heavily scrutinized, or just thrown out for being too much work to review.
As to whether it would be the "equivalent" or a PhD, it won't get through a resume parser, and won't get you a teaching position or even a postdoctorate position, but companies probably won't care.
Thanks for the advice, I appreciate it’s clarity. I guess I’d be better served through a PhD program anyway, for the doors opened when it comes to publishing etc
PhD by publication is usually an (unofficial, unadvertised) option in many UK/Australian universities. If you published four or five good papers, that would more than fulfil the requirements for the degree (if the university technically requires you to write a thesis, at that point you just bind your four papers together and call it a thesis).
At least in Australia, you need to begin a PhD before, or at least very close to, when you start working on this stuff. So if you go this route, make sure you talk to an interested professor+university before you go off for a few years and work on stuff. The reason being is that universities get a large award from the government when students complete a PhD, and so all of the work you publish has to be “part of” your degree - you can’t bring a large amount of work along from elsewhere.
I understand the situation might be different in the US, but in the UK and in other countries a PhD with a hands-off supervisor can be as lonely as your heart desires.
I'm actually looking into UK PhD programs, especially since I already am recieving a joint masters in CS and Biology
I think you need to tell yourself if you need to get PhD degree, or like to have that capability people can do as PhD. Just do it if you just need that capability, and I bet you can learning more that you expected.
PhD student here doing neuromorphic computing and running deep learning models on analog hardware. The 'neuromorphic computing" is a vaguely defined term, which can mean anything from developing novel Spiking NN models to the design of hardware capable to run those models (e.g. TrueNorth). Wetware computing is probably even lesser understood area. Neither has much to do with deep learning field, which does not use spiking NN computation.
On the other hand, there are very promising research directions precisely in the intersection of neuroscience and deep learning (e.g. biologically plausible credit assignment problem, learning algorithms of the brain, objective functions in the brain, sensori-motoric integration, etc). It's rare to find a person with solid understanding of both fields, especially in academia.
If I were you I would look at residency programs at big companies (Google Brain, Facebook AI fellowship, etc). Also, there's a company called Numenta which is probably looking for people with your skills (they are trying to reverse engineer neocortex' algorithms). From what I heard, those places give you some freedom to choose your own problems, encourage publishing papers, and provide very bright people as mentors.
There would be several professors interested in students like you. You have to mail them and talk about your topic and idea. There are many who would accept an enthusiastic Ph.D.
PhD by Publication is an option here in the UK. Check this link from Warwick University to see if you're eligible and if this is something you'd be interested in: https://warwick.ac.uk/study/postgraduate/research/phdbypubli...
Interesting! Does this mean that you can focus on publishing a few papers and make a proposal highlighting all your contributions and papers, then apply to get a PhD title.
Nope, you still formally sign up like a standard PhD student but will organize your thesis as a series of published and peer-reviewed papers, instead of the standard format, at the normal end of your PhD program. There is neither an accelerated path nor a diy option because the PhD is strictly meant as training time in the UK, much more like an apprenticeship than autonomous contribution.