HTM core detector pandaVis - Bugfix distal synapses

[Zbysekz]

Small bugfix, as result clicking on cells was causing exception

[Zbysekz]

Second addition, while this PR is already opened,

I discovered out big problems with order calls, that i have in all scripts (2D object recognition, hotgym example and this)

It spend literally hours to get the execution order right in my head and i must say that it was very hard. :tired_face:

But finally this is the right. I changed here the order of calls. Basically

1. for first timestep get predictions
2. calculate activeCells (winners..)
3. get predictions for next time step
4. Update panda data and block execution
5. go to 2.

[breznak]

@Zbysekz can we continue your work here? Either by merging as is, or by revamping the API?

[Zbysekz]

@Zbysekz can we continue your work here? Either by merging as is, or by revamping the API?

I am now working on the new interface, baking data to the database. (already got working partially) This then would be much simplier, i am for cancelling this PR and then i start new one with the revamp when it will be ready.

[breznak]

I am now working on the new interface,

oh, that's great :+1: , I was worried the progress got lost.

Could you post an example (pseudocode) how the new API would be used in 3rd party? As in here?

[Zbysekz]

I think it will be like this: Method to define system structure same as now:

• BuildPandaSystem()

global variable for "pandaBaker" with database file specified pandaBaker = PandaBaker('/home/zz/Desktop/test1.db')

Filling data into dicts

        # fill up values
        pandaBaker.inputs["SL_Consumption"].stringValue = "consumption: {:.2f}".format(consumption)
        pandaBaker.inputs["SL_Consumption"].bits = consumptionBits.sparse

        pandaBaker.inputs["SL_TimeOfDay"].stringValue = record[0]
        pandaBaker.inputs["SL_TimeOfDay"].bits = dateBits.sparse

        pandaBaker.layers["SensoryLayer"].activeColumns = activeColumns.sparse
        pandaBaker.layers["SensoryLayer"].winnerCells = tm.getWinnerCells().sparse
        pandaBaker.layers["SensoryLayer"].predictiveCells = tm.getPredictiveCells().sparse
        pandaBaker.layers["SensoryLayer"].activeCells = tm.getActiveCells().sparse

And custom data for plots user configurable window like:

pandaBaker.customStream["myTagNameOfStream"] = value

I will probably use exactly this universality for raw anomaly: pandaBaker.customStream["SensoryLayer_rawAnomaly"] = rawAnomaly

I will try to make this as simple as possible