Deep analog learning able to function like a human brain? Indeed, it seems to be the case, as all of this is the answer to the question many are asking about machine learning: will it be possible to increase their processing power of the results? Thanks to this brain, and by using electronic devices similar to neurons, it is possible to have this enhancement capable of making machine learning much more efficient.
Not only that, because the proposed model has been significantly improved by Murat Onen of MIT and its team to make this brain even faster. This decision was taken after finding that the proposed solution it wasn’t quite as fast as expected and this has led to new actions to be carried out on it. All thanks to a new electrolyte that has definitely changed the cards on the table.
Analogue deep learning can learn faster with the new element
What will all this be for? What are the plans for the future? Certainly, at the base, there will be a massive study of the components of this analog brain which, as already explained, mimics the connections between forming neurons the deep neural networks. This process is given more thanks to the programmable proton resistors, components of the analog brain, able to be established by man through training algorithms. This will allow you to always discover new features, which will lead to the resizing of the same for possible insertion in other small devices.
“Once we get the analog processor, we won’t be training networks that everyone currently works on. Instead, they will be of unprecedented complexity that no one can afford, and therefore that will surpass all the others. In other words, this is not a faster car, this is a spaceship “ he has declared Murat Onen on the new analog deep learning.
#Analogue #deep #learning #heres #hardware #news