補間関数マシンとしてのAI実験 AI Test as an Interpolation Function Machine
ニューラルネットは補間関数だと前に述べた(https://randomwalkjapan.blogspot.com/2019/05/ai-reality-of-booming-ai.html)が、下の自作ツールで試してみよう!Ealier I mentioned that neural net is an interpolation function(https://randomwalkjapan.blogspot.com/2019/05/ai-reality-of-booming-ai.html) so let's try it out in below tool I made!
10個のデータポイントを色々変えると、関数が難しくなるにつれて、ニューロン数、繰り返し回数、中間層の数を増やさないと追従できなくなることが確認できる。If you change 10 data points, as function become complex, you can confirm that interpolation becomes difficult unless you increase number of neurons, epochs, layers.
Program is at
https://github.com/tanakayutaka/Tensorflow.js-tools-/upload/master .