# Prediction for One Hot Gym - Large swarm

**URL:** <https://discourse.numenta.org/t/prediction-for-one-hot-gym-large-swarm/6209>\
**Category:** NuPIC\
**Created:** [June 27, 2019, 2:09pm UTC](https://discourse.numenta.org/t/prediction-for-one-hot-gym-large-swarm/6209 "2019-06-27T14:09:22Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![helena\_Thielen](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/helena_thielen/32/4343_2.png) [@helena\_Thielen](https://discourse.numenta.org/u/helena_Thielen)\
**Post date:** [June 27, 2019, 2:09pm UTC](https://discourse.numenta.org/t/prediction-for-one-hot-gym-large-swarm/6209/1 "2019-06-27T14:09:22Z")

</div>

Hey everybody 🤗👋

I am working with the example on

> **[numenta/nupic](https://github.com/numenta/nupic/tree/master/examples/opf/clients/hotgym/prediction/one_gym)**
>
> Numenta Platform for Intelligent Computing is an implementation of Hierarchical Temporal Memory (HTM), a theory of intelligence based strictly on the neuroscience of the neocortex. - numenta/nupic

Everything worked fine so far…I could reproduce a result, that was quite similar to the one Matt showed on YouTube ([https://www.youtube.com/watch?v=S-0thrzOHTc](https://www.youtube.com/watch?v=S-0thrzOHTc)). 🤭🤩

Now I wanted to investigate the influence, the swarming has on the result.  
So I ran a swarm with “swarmSize”: “large”  
Now I did the Prediction with the resulting model\_params.  
And actually… I think it looks too good… Like an over fitting. 🙄🥺

Here one can see (a part of) the result:

 ![Figure_1](https://canada1.discourse-cdn.com/flex030/uploads/numenta/original/2X/5/5f95b020f7f1d3ee04435c1c7b19a119851581b8.png)

The Prediction and the Data are most of the time exactly the same (just sometimes there is a little difference).  
Now my question: is this normal when working with a large swarm? Is this the reason why we normally take a medium one?

Thank you all very much in advance for your help 🤗😍🙃

---

<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:** [June 27, 2019, 3:48pm UTC](https://discourse.numenta.org/t/prediction-for-one-hot-gym-large-swarm/6209/2 "2019-06-27T15:48:01Z")

</div>

Did you compare the model params the large swarm returned vs the medium? there may not be much of a different. A large swarm sometimes won’t even help much.

Also, if you are plotting the data, you want to use the [inference shifter](https://discourse.numenta.org/t/issue-using-nupic-anomaly-output/4809/2) to align the predictions with the observations. If predictions are trailing exactly one time step behind observations, it means the system does not have a valid prediction, so it will simply return the value it just observed.

---

<div class="post-metadata">

**Author:** ![helena\_Thielen](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/helena_thielen/32/4343_2.png) [@helena\_Thielen](https://discourse.numenta.org/u/helena_Thielen)\
**Post date:** [July 11, 2019, 8:21am UTC](https://discourse.numenta.org/t/prediction-for-one-hot-gym-large-swarm/6209/3 "2019-07-11T08:21:19Z")

</div>

sooo first sorry for the late reply.  
I was super sick the last 2 weeks.

ok I just misunderstood the thing with the shift. Super big sorry for that.

Anyway, I just ran the large swarm again and afterwards the prediction with the MODEL\_PARAMS from the large swarm. And finally it worked. 🥳😁🤩  
I don’t know what happened last time. 😅

Soo, thank you very much for the help 🤗