flâneur

Will

9 followers · 6 following · 1490 views

on the atlas — 55

highlights — 3

  • Hopfield networks serve as content-addressable ("associative") memory systems with binary threshold nodes, or with continuous variables.[6] Hopfield networks also provide a model for understanding human memory.[7][8]
    Hopfield network
  • The Ising model of a recurrent neural network as a learning memory model was first proposed by Shun'ichi Amari in 1972[2][3] and then by William A. Little in 1974,[4] who was acknowledged by Hopfield in his 1982 paper.[1] Networks with continuous dynamics were developed by Hopfield in his 1984 paper.[6] A major advance in memory storage capacity was developed by Krotov and Hopfield in 2016[9] through a change in network dynamics and energy function. This idea was further extended by Demircigil and collaborators in 2017.[10] The continuous dynamics of large memory capacity models was developed …
    Hopfield network
  • One version of this question—reminiscent of adversarial examples in neural networks—is just to ask what changes need to be made to an initial condition to “flip its result”. Or, put another way: let’s say one has a system (like the GKL rule) that basically achieves correct consensus for almost all randomly chosen initial conditions. Now we ask the question of whether there is a systematic way to tweak a given randomly chosen initial condition to make it “lead to the wrong answer”. (One can think of this as a bit like asking whether one can find a nonce that will make a cryptographic hash come …
    The Problem of Distributed Consensus—Stephen Wolfram Writings