1

Mastering my Machine Learning, I try arithmetic:

ClearAll["Global`*"];
trainingset = {"2+2" -> 4, "2+3*2" -> 8, "(12+7)*5" -> 95, "7*6" -> 42,
 "7+22" -> 29, "4+5" -> 9, "4*1+5" -> 9, "17*3+4*5" -> 41, "7+9*2" -> 25, "11+3" -> 14, 
"6+6" -> 12, "4*5+6" -> 26, "5*7" -> 35, "3*2" -> 6, "3+2" -> 5, "9*3" -> 27, 
   "3*9" -> 27, "6*3+8*2" -> 34, "5*4" -> 20};

Making use of different methods, I'd like to compare the results:

p = Predict[trainingset, Method -> "RandomForest"]; p["4+6"]

24.6316 

r = Predict[trainingset, Method -> "LinearRegression"]; r["4+6"]

20.3651

t = Predict[trainingset, Method -> "NearestNeighbors"]; t["4+6"]

26.

u = Predict[trainingset, Method -> "DecisionTree"]; u["4+6"]

14.1

w = Predict[trainingset, Method -> "GradientBoostedTrees"]; w["4+6"]

13.7925

However,

s = Predict[trainingset, Method -> "NeuralNetwork"]; s["4+6"]

Beep:The kernel Local has quit (exited) during the course of an evaluation.

Looking in the help, I don't understand that behavior of Predict.

Addition. I mean the result with Method -> "NeuralNetwork". It's clear that a training set of 19 elements is too small to train Predict. My another question is how to generate a training set consisting of 200 examples in an automatic way.

user64494
  • 26,149
  • 4
  • 27
  • 56
  • 1
    Works fine in version 11.3 on macOS (apart from the fact that result of "4+6" is rubbish). Probably a nasty memory related bug. Try also to restart the kernel. Also you can go into the help menu and click "Why the beep?..." Maybe that gives you some idea what happened. Maybe not. – Henrik Schumacher Jul 24 '18 at 17:19
  • @Henrik Schumacher: Thank you for your interest to my question. If you carefully read my answer, you will see I looked in "Why Beep". A new kernel does not help. – user64494 Jul 24 '18 at 17:37
  • Ah sorry. Well, then I guess this is a OS-related bug. – Henrik Schumacher Jul 24 '18 at 17:39
  • @Henrik Schumacher: I work in version 11.3 on Windows 10 32-bit. – user64494 Jul 24 '18 at 17:41
  • It works fine on my machine running Windows 10, 64bit – Niki Estner Jul 24 '18 at 17:48
  • @Niki Estner: Thank you for good news. Can you kindly tell us the obtained result? TIA. – user64494 Jul 24 '18 at 17:51
  • You get a different result every time you run it with a fresh kernel. Usually around 25, because that's the mean of the training values. You realize that this isn't going to work, right? – Niki Estner Jul 24 '18 at 17:56
  • @Niki Estner: Yes. I hope I am clear in my question. The command quits at the first example of 19. – user64494 Jul 24 '18 at 18:03
  • @user64494 I think it's time to contact support about it. – Henrik Schumacher Jul 25 '18 at 06:51
  • @Henrik Schumacher: I will think about it, but the deficiency is minor. How about my another question: how to generate a training set to this end in an automatic way? – user64494 Jul 25 '18 at 07:00

0 Answers0