Closed hadarszostak closed 7 years ago
The "dimensions" in the GRT correspond to the number of dimensions in your input data (aka, the number of features in your input data or the number of attributes).
If you have a single-axis accelerometer, then you might have 1 feature in your data. If you have a three-axis accelerometer (e.g., one that gives you X, Y, Z), then you would have 3 dimensions in your data.
This example shows you how to set the number of dimensions in your data before you add any new data:
You mentioned that you have 101 below, this sounds more like 101 samples (e.g., examples, data points) as opposed to features). Are you trying to recognize some time series signal in your accelerometer data, or just recognize a basic orientation?
The 101 samples are time series. For my understanding now, I can't add a vector (I can add multiple dimensions results, but not multiple results for each dimension), I need to use addSample with each received sample ( of the 101 ).
Edit: Thanks, it works now.
I see at the examples that Dimensions always set to 1. Even in getting started page,
but I use 101 samples (Accelerometer x axis 101 samples) to train the pipeline, trainingData.addSample((UINT) 1 , DataVecot) results with : [WARNING ClassificationData] addSample(const UINT classLabel, VectorFloat &sample) - the size of the new sample (101) does not match the number of dimensions of the dataset (1), setting dimensionality to: 1 [ERROR] process(const VectorFloat &inputVector) - The size of the inputVector (101) does not match that of the filter (1)!
Help?
Thanks