'Memtransistor' Forms Foundational Circuit Element to Neuromorphic Computing

Re:'Memtransistor' Forms Foundational Circuit Element to Neu... by jimrandomh on Monday February 26, 2018 @09:53PM (#16470768)

All of the MNIST accuracy scores mentioned in these links are terrible, even by the standards of untuned novelty algorithms.

[ Parent | on HN ]

Re:'Memtransistor' Forms Foundational Circuit Element to Neu... by cs702 on Tuesday February 27, 2018 @08:37AM (#16473179)

Of course, because the nets we're talking about are really, really tiny.

As you can see for yourself in figure [1]c, the memristor net used in that paper has only 15 units (yes, fifteen): five memristors in the "reservoir" and 10 output neurons. The entire net is one linear transformation followed by a SoftMax, with a total of 5 × 10 = 50 parameters.

It's an impressive result that only hints at the things that will be possible when we have nets with millions or billions of memristor units.

[1] https://www.nature.com/articles/s41467-017-02337-y/figures/1

[ Parent | on HN ]

Re:'Memtransistor' Forms Foundational Circuit Element to Neu... by jimrandomh on Tuesday February 27, 2018 @06:24PM (#16479116)

I don't think it says anything about what a million or billion memristor network would be able to do. I don't think they scale in the same way normal machine learning algorithms do.

[ Parent | on HN ]