Neural Turing Machines

Neural Turing Machines give a neural controller a separate, mutable memory and make its read and write addresses differentiable. Read it to see how content lookup and relative shifts combine, why the copy trace looks like a learned algorithm, and where the paper's preliminary synthetic evidence stops.

Reading focus: How a normalized weighting turns a discrete memory address into a trainable soft read or write. Why content addressing finds a matching row while location addressing iterates from the previous focus. What the copy memory trace establishes, and why fixed capacity and missing seed variation bound the claim.

arXiv 2014. Alex Graves, Greg Wayne, and Ivo Danihelka. 28 min read, medium difficulty.