Should the TM predict the pattern perfectly accurately?

Hello, simple qustion in topic title.

For example, its t-1 prediction:
image

And t0 active columns state:
image

One column predicted wrong.
Sometimes there may be 2-3 incorrectly predicted columns.
Its valid, or not?

1 Like

The accuracy of TM predictions generally depends on how much ambiguity and noise exists in the sequences.

So this scenario is perfectly plausible, except in scenarios of noiseless simple patterns – which should have 0 unpredicted columns given enough repetitions.

2 Likes