# Implementing SDR long term memory?

**URL:** <https://discourse.numenta.org/t/implementing-sdr-long-term-memory/8274>\
**Category:** Implementations\
**Tags:** memory\
**Created:** [February 26, 2021, 6:12pm UTC](https://discourse.numenta.org/t/implementing-sdr-long-term-memory/8274 "2021-02-26T18:12:47Z")\
**Posts on this page:** 1\
**Showing post:** 5

<div class="post-metadata">

**Author:** ![marty1885](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/marty1885/32/1891_2.png) [@marty1885](https://discourse.numenta.org/u/marty1885)\
**Post date:** [February 27, 2021, 5:44am UTC](https://discourse.numenta.org/t/implementing-sdr-long-term-memory/8274/5 "2021-02-27T05:44:00Z")

</div>

> [@mraptor](#):
>
> what about point 3 ? any idea

TM can recall even with noise/loss data. So I think it’ll work (unless the input is so corrupted that it is unrecognizable).

I’ve did this in my [RL project](https://discourse.numenta.org/t/hierarchical-temporal-memory-agent-in-standard-reinforcement-learning-environment/7113) to force a TM to predict future motor commands in the absent of DT-errors. This is a very hacky way of doing it. TM is not meant to be used as a KV-store. But it can be forced into one. Though I don’t know how well it works quantitatively.

---

_[View the full topic](https://discourse.numenta.org/t/implementing-sdr-long-term-memory/8274)._