On Hopfield Networks. From the overall mannequin to a particular… | by Arun Jagota | Oct, 2024

From the overall mannequin to a particular case

Photograph by Shubham Dhage on Unsplash

The 2024 Nobel Prize in Physics went to John Hopfield and Geoffrey Hinton, John Hopfield on his work on the so-called Hopfield Community.

A couple of many years again, my PhD thesis was on Hopfield-style Networks. I figured this was a very good time to put in writing this publish.

I’ll give attention to Hopfield networks. I’ll start with the only instance and illustrate the way it works, then carry within the idea of an power perform and minimizing it regionally, then cowl the particular case launched in my PhD thesis, then come again to the extra basic case and talk about associative recollections and optimization use circumstances, then come again to the particular case to instantiate these purposes in a extra particular setting.

Let’s start at the start. First off, what’s a Hopfield Community?

Contemplate the image under.

Interpret it as follows. We have now two neurons, every can tackle a worth of 1 (suppose “firing”) or -1 (suppose “not firing”). The 2 neurons are related by a optimistic weight, denoting a synapse of optimistic affect. Which means that, from the angle of any one of many neurons, say the primary one, it needs the opposite one to be in the identical state as it’s — “firing” or “not firing”.