Anomaly detection error: TypeError: unsupported operand type(s) for +=: 'float' and 'NoneType'

[tjdurant]

Hello,

I am running through the tutorial from (https://github.com/numenta/nupic/tree/master/examples/opf/clients/hotgym/anomaly/one_gym). I am getting the error below. I am running CentOS7, Python 2.7.13 :: Anaconda 4.3.0 (64-bit),

$ python run_anomaly.py

Gives this output:

Starts a NuPIC model from the model params returned by the swarm
and pushes each line of input from the gym into the model. Results
are written to an output file (default) or plotted dynamically if
the --plot option is specified.

Creating model from rec-center-hourly...
Importing model params from model_params.rec_center_hourly_model_params
Preparing to output rec-center-hourly data to rec-center-hourly_out.csv
Read 100 lines...
Read 200 lines...
Read 300 lines...
Traceback (most recent call last):
  File "run_anomaly.py", line 151, in <module>
    runModel(GYM_NAME, plot=plot)
  File "run_anomaly.py", line 141, in runModel
    runIoThroughNupic(inputData, model, gymName, plot)
  File "run_anomaly.py", line 122, in runIoThroughNupic
    output.write(timestamp, consumption, prediction, anomalyScore)
  File "/root/nupic/swarm/nupic_anomaly_output.py", line 93, in write
    value, anomalyScore, timestamp
  File "/root/anaconda2/lib/python2.7/site-packages/nupic/algorithms/anomaly_likelihood.py", line 386, in anomalyProbability
    skipRecords=numSkipRecords)
  File "/root/anaconda2/lib/python2.7/site-packages/nupic/algorithms/anomaly_likelihood.py", line 460, in estimateAnomalyLikelihoods
    verbosity = verbosity)
  File "/root/anaconda2/lib/python2.7/site-packages/nupic/algorithms/anomaly_likelihood.py", line 672, in _anomalyScoreMovingAverage
    MovingAverage.compute(historicalValues, total, record[2], windowSize)
  File "/root/anaconda2/lib/python2.7/site-packages/nupic/utils.py", line 81, in compute
    total += newVal
TypeError: unsupported operand type(s) for +=: 'float' and 'NoneType'

[rhyolight]

Someone else just reported this in https://github.com/numenta/nupic/issues/3774.

You are using Anaconda? (If so, we don't support it yet, but you can help).

[tjdurant]

Yea, we're using Anaconda, but we use pip to install it into the virtual environment:

$ conda list

nupic                     1.0.3                     <pip>
nupic.bindings            1.0.0                     <pip>

I saw on #3774 that he is able to get anomaly scores, but I don't even get those. I'm not entirely sure about where my problem is in relation to his, so I'll defer to you if you think ours are sufficiently similar that I can just follow #3774 for a solution.

Thanks! Thomas

[rhyolight]

I don't trust NuPIC on Anaconda because we don't have a CI build for that environment yet. So I would just assume it will not work properly on Anaconda at this point. Will defer further tech discussion to #3774. Thanks for the report.