I entered 374 data results (yes, by hand, on that grid!!) to try to get a prediction.
It takes 2:34 for the beast to load on my ancient CLIE S320 with that amount of data.
It seems some vector files broke and I kept getting a wrong prediction. (It takes about 3 minutes for it to predict on my sloooow CLIE!)
With my initial ~35 test entries, the prediction I got was screwy — it was recursive! — but after examing the source data, I could see why it came up with that prediction.
The program author suggests I try M-5 Neural Network Computer 1.0 instead. Agh, what a beast!
I’m also wondering what the desktop version of the Numenta software would do…
That’s all for this weekend.
Previously in this blog:
You Can Now Play With A Hawkins Machine!