Here’s a nice video about interview of Geoffrey Hilton on his new Forwarad Forward algorithm, which I am currently working on based on his preliminary research. This interview is pretty cool as it shows us how a great idea was generated: thousands of times of trying, large amount of reading, consistent curiosity and explorations… and also hard working.

When talking about reservoir computing, one must mention that the reservoir’s main requirements are nonlinearity and a memory effect, also known as the ‘echo state property’ in the context of Echo State Networks (see more in one of my other blog: (What is and why reservoir computing). But why it isn’t a system has certain memory effect must also have nonlinearity? Intuitively, Nonlinear systems often exhibit complex behaviors that are sensitive to initial conditions, such as chaos. Since memory effects introduce a form of complexity— where the output depends not just on the current input but also on past inputs— it can be easy to conflate this complexity with nonlinearity.

To creat a repository of an exsisting project in github from your local server:

The beginning of the story

Micromagnetic simulations

Hey!