-
How to write a thesis in a month
I just finished the initial submission of my PhD thesis, which I started writing on Jan. 7th. It was a tight fit; although it’s based on 3 published manuscripts, it still required writing a broad intro and conclusion (about 16k new words total) plus a lot of reformatting to make it into a cohesive whole.
-
A solution for bad teaching
Excellent article over at the NY Times on the drawbacks of having researchers in teaching positions. There’s a lot of research showing that the relationship between research ability and teaching ability is very close to 0. So why are we making researchers teach? Shouldn’t teachers teach and researchers research? The result, then, is that more
-
Postdoc positions at Columbia in statistical neuro and machine learning
Just got an email from Dr. Paninski. There are two postdoc positions available at the new Grossman center in statistical neuroscience and machine learning. This is an amazing opportunity if you’re looking for a theory-oriented researcher. Liam’s recent research output is impressive, he’s super smart, a very nice guy, and he has a kick-ass beard.
-
Theano: numerical computation in Python
Theano is a very interesting numeric library for Python that I covered briefly a few years ago. Coming from the machine learning group at Université de Montréal – i.e. Yoshua Bengio et al. – it is well adapted to the kinds of numerical tasks that frequently occur in machine learning problems, in particular deep neural
-
Adagrad – eliminating learning rates in stochastic gradient descent
Earlier, I discussed how I had no luck using second-order optimization methods on a convolutional neural net fitting problem, and some of the reasons why stochastic gradient descent works well on this class of problems. Stochastic gradient descent is not a plug-and-play optimization algorithm; it requires messing around with the step size hyperparameter, forcing you
-
Workshop on statistical challenges in neuroscience
My old buddy Simon Barthelmé as well as Nicolas Chopin and Adam Johansen are organizing a 3-day workshop in Warwick, UK in September on statistical challenges in neuroscience. Some of the invited speakers include J.D. Victor, Robert Kass and Aapo Hyvärinen. Great chance to discuss and present your latest methods-heavy research. Here’s the pitch: Workshop
xcorr: AI & neuro 