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 o

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