M. Jordan's Post on Deep Learning

I just recently came across this great post from M. Jordan. He acknowledges neuroscience as being one of the major areas of research "for the next few hundred years" but points out that the major advances in theory and conceptual thinking about learning and inference are not really coming from that field.

While this is in fact one reason why I personally left this field, I like to jump in and defend a bit: Trying to understand how the brain (and the body it commands) is and will be for a very long time one of the major scientific challenges. Working out brain theories (and this is what "Computational Neuroscience" is ultimately doing) will involve both the "wet parts" from the neuroscience labs, the dryer part from cognitive neuroscience and psychology, and the really dry parts of machine learning and learning theory, and of course, all the rebranded models from physicists and applied mathematicians. Even approaches to "solve intelligence" without the burden of being compatible with neuroscience findings will ultimately have to serve a scientific theory of how the mind and neurons relate. This is not big news. It was always the driving force of early neural networks, cybernetics, and the like.

Personally, I just like to add that keeping the fields of the neurosciences, neural networks, learning theory, machine learning, applied math, etc. not too separate may be important, but in the real-world it is often too frustrating.

If you want to get a bigger picture, check out the upcoming Springer Encyclopedia on Computational Neuroscience, where I was the editor of the section on "Cortex: Computation and Modeling" with aprox. 15+ articles from leading scientist in the field. In my introduction I sketch such a picture in a bit greater detail.

Program for the Berlin buzz words 2014 conference is available.

Check out the program of the Berlin buzz words, which is now out: berlinbuzzwords.de.

My very first post

This is my very first post in this blog. I had various blogs using Wordpress before, but now I am ready to enjoy Google's platform. It is a natural move, though, since I finally surrendered to Google.