NanoNets : How to use Deep Learning when you have Limited Data
Sarthak Jain
1.1K15

Interesting piece. However, advertising a 100% accuracy on a test set with two samples is very dangerous and should have 100 caveats. A randomly initialized model can have a 25 % chance of getting a 100 % test accuracy on a binary task with only two examples. No conclusions should be drawn from a test set of size 2. period.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.