Consider allowing some prediction evidence (k*c) to bleed into current event with projection.
OpenNARS treatement to simply revise doesn't work well, but with a small k (say 0.3) it should be fine. Similar idea as in Kalman filter, if prediction is believed too much over observation it won't work well.
patham9
commented
2 years ago
Consider allowing some prediction evidence (k*c) to bleed into current event with projection. OpenNARS treatement to simply revise doesn't work well, but with a small k (say 0.3) it should be fine. Similar idea as in Kalman filter, if prediction is believed too much over observation it won't work well.