This model was used to generate the top scoring submission to the 2017 Parkinson's Disease Digital Biomarker DREAM Challenge, Subchallenge 1 (https://www.synapse.org/#!Synapse:syn8717496/wiki/422884).

It is recommended to first store the result of load_guanlab_model to an R object, then pass that object to the model parameter. Otherwise it is necessary to load the weights into the model each time this function is called.

guanlab_model(sensor_data, models = load_guanlab_model)

Arguments

sensor_data	An n x 4 data frame with columns t, x, y, z containing kinematic sensor (accelerometer or gyroscope) measurements. Here n is the total number of measurements, t is the timestamp of each measurement, and x, y and z are linear axial measurements.
models	A list of models to use for prediction.

Value

10 different "features", which are actually just predictions generated by the same neural net architecture trained to different local minima.

Details

This model may be used to generate some robust predictions to distinguish Parkinson's patients from controls.