# Benchmarks/capabilities

**URL:** <https://discourse.numenta.org/t/benchmarks-capabilities/6920>\
**Category:** Applications\
**Created:** [December 15, 2019, 2:26pm UTC](https://discourse.numenta.org/t/benchmarks-capabilities/6920 "2019-12-15T14:26:48Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![ConnectedRegions](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/connectedregions/32/5487_2.png) [@ConnectedRegions](https://discourse.numenta.org/u/ConnectedRegions)\
**Post date:** [December 15, 2019, 2:26pm UTC](https://discourse.numenta.org/t/benchmarks-capabilities/6920/1 "2019-12-15T14:26:48Z")

</div>

When talking to the general AI/deep learning community, one of the things they always emphasize is benchmarks.

I saw Numenta’s applications page - [https://numenta.com/machine-intelligence-technology/applications/](https://numenta.com/machine-intelligence-technology/applications/)  
Where HTM can be applied to analyze human behavior in, predict energy usage, detect analomies in the behavior of moving objects, etc

How there been any benchmarks that compare Numenta’s HTM to deep neural networks when it comes to these kinds of tasks?

Because HTM is based off the design of the neocortex, are there additional capabilities it has shown that traditional deep neural networks are not capable of today?

---

<div class="post-metadata">

**Author:** ![rhyolight](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/rhyolight/32/3922_2.png) [@rhyolight](https://discourse.numenta.org/u/rhyolight)\
**Post date:** [December 16, 2019, 10:31pm UTC](https://discourse.numenta.org/t/benchmarks-capabilities/6920/2 "2019-12-16T22:31:30Z")

</div>

> [@ConnectedRegions](#):
>
> How there been any benchmarks that compare Numenta’s HTM to deep neural networks when it comes to these kinds of tasks?

I hope you have seen the [Numenta Anomaly Benchmark](https://github.com/numenta/NAB)?

> [@ConnectedRegions](#):
>
> Because HTM is based off the design of the neocortex, are there additional capabilities it has shown that traditional deep neural networks are not capable of today?

HTM systems naturally model sequences through time, which deep learning and gradient descent have a harder time with.