Revitalize Numenta HTM - Hierarchical Temporal Memory
Numenta HTM (Hierarchical Temporal Memory) is a biologically constrained machine intelligence technology that models the structural and algorithmic properties of the mammalian neocortex
https://github.com/berlinbrown/htm.java
This is adding updates to the project.
The visualizer simulates a room thermostat reporting 24 hourly readings each day:
```text
12 AM 64°F → 1 AM 63°F → ... → Noon 76°F → ... → 11 PM 65°F → next day
```
This example runs the complete learning path:
```text
hour ─────────→ periodic ScalarEncoder ─┐
├→ Spatial Pooler → active columns → Temporal Memory
temperature ──→ ScalarEncoder ─────────┘
```
The temperatures are simulated sensor readings, but they are ordinary real-world
values processed by the actual encoder and Spatial Pooler. Numbers such as
`[2, 3, 5]` are internal column addresses, not temperatures and not numbers the
model is learning to count.
Also converted some of NAB - https://github.com/numenta/NAB/
Let's go AGI!
Comments