Continuous Online Sequence Learning with an Unsupervised Neural Network Model
Created on 2021-01-31T13:51:17-06:00
Reiteration of HTM theory.
Robustness to signal noise, sparseness.
Temporal memory is not robust to temporal noise.
HTM applied to grammar learning was 98.4% accurate while LSTMs were 100%
- --
I don't consider being ~2% less than LSTM to be a considerable flaw given the system is able to learn in real time and LSTMs are not.