Anticipation investigation

[patham9]

[patham9]

This PR won't make it, but for future reference, Anticipation was the approach I tested heavily against ANSNA's feedback handling. A summary from our related discussions:

The key feature of any prediction feedback handling, applying to both Anticipation and Assumption of Failure:

1. Truth expectation of the hypothesis should increase when successful and decrease when unsuccessful.
2. Truth expectation should become more stable over time (comes for free, thanks to confidence increase in Revision)

Anticipation additionally features: 3. Anticipation doesn't add negative evidence at all if the hypothesis turns out to be successful.

Point 3 is nice theoretically. Practically it does not matter, because neither decision making nor prediction reliability relies on 3. they relie on 1 and 2 and only on 1 and 2.

Disadvantage of Anticipation:

• Effort to store anticipations and to match them to input until they expire
• Difficulty to make a proper deadline estimate. Advantage:
• Frequency can reflect success rate better.

For OpenNARS: No need to change to Assumption of Failure, as it would be a change without much gain. For ANSNA: Assumption of Failure is simpler, and has same effectiveness, hence ANSNA will stay with it. For ALANN: I suggest to also go with Assumption of Failure due to its simplicity.