Re:Yann LeCun on the IBM neural net chip
by JoeAltmaier on Sunday August 10, 2014 @10:02AM
(#8159726)
Bad science? Trying things? That's fundamentally what true science is all about. Experimentation is where theories are supposed to come from. Remember Nature just put a ton of neurons together to see if it did anything.
What would be called 'good science'? Reading articles and spinning tales about what comes next? Regurgitating summaries of others' work?
This money would be better spent in experiments to find out how neurons work. IF neurons are well studied, and large scale models of the brain using IF models have been done before[1]. The result? Nothing.
It would be an experiment if they were testing a new model. This is a simulation.
[1] http://www.izhikevich.org/human_brain_simulation/Blue_Brain.... of Large-Scale Brain Models
Re:Yann LeCun on the IBM neural net chip
by JoeAltmaier on Sunday August 10, 2014 @10:44AM
(#8159840)
This experiment was different in some way? More 'neurons' this time? That qualifies as an experiment.
This experiment was different and inferior in almost all ways. Both less neurons and simpler neurons than simulated before. We had simulated 100x more neurons in more biological detail than what TrueNorth team did.
This is a great asynchronous circuit power efficiency research, and not a neuroscience research at all.
The fact that Truenorth can learn approximations is not really surprising, we know that thresholded units can approximate well[1]. They should have implemented compartmental neurons [2]
[1] http://en.wikipedia.org/wiki/Universal_approximation_theorem [2] http://en.wikipedia.org/wiki/Compartmental_modelling_of_dend...