SeqReg Module

function LoadDataFromFolder(…)

Parameters:

  • folderpath: path to location of saved data (in specified csv format)

  • xname: name of the csv column with xdata

  • yname: name of the csv column with ydata

  • tname: name of the csv column with time data (defaults to Time)

  • datatype: type of data in the x column (either Value, txtFilePath, PCAnpy) (defaults to Value)

  • pcskeep: amount of pcs to keep if using data from PCA data reduction (either and interger or “ALL”) (defaults to ALL)

Output:

  • xds: list of all loaded x data of shape (# of experiments, -)

  • yds: list of all loaded y data of shape (# of experiments, -)

  • timeds: list of all loaded time data of shape (# of experiments, -)

function PrepareData(…)

Parameters:

  • x: list of all loaded x data (xds output from LoadDataFromFolder function)

  • y: list of all loaded y data (yds output from LoadDataFromFolder function)

  • time: list of all loaded time data (timeds output from LoadDataFromFolder function)

  • seqlen: length of sequences to generate using a rolling sampling method

  • stride: stride to use for generating sequences using a rolling sampling method

  • dt: timestep between each data point. This is only relevant if using the FFT

  • fft: a boolean defining whether to use the fft to transform x data sequences to the frequency domain (defaults to False)

  • seqout: a boolean defining whether the y data should be a single value or sequence for outputs

Output:

  • x1: numpy array of transformed x data

  • y1: numpy array of transformed y data

  • t1: numpy array of transformed time data

function Model(…)

Parameters:

  • modelname: the name of the model architecture to be used (HydReg, Hit2Flux, ImgReg)

  • savemodelpath: either the path to the location of saved weights if train=False or path to location where weights will be saved if train=True

  • train: boolean specifying whether a model is to be trained or using a pretrained model (defaults to False)

  • xtrain: numpy array containing training x data prepared based on PrepareData function (defaults to None)

  • ytrain: numpy array containing training y data prepared based on PrepareData function (defaults to None)

Output:

  • model: a tensorflow or sklearn model

  • if train=True saved weigths or model to specified path

function Analyze(…)

Parameters:

  • model: tensorflow or sklearn model

  • savepath: path to location where metrics and plots should be saved

  • xtest: numpy array of x test data generated from PrepareData function

  • ytest: numpy array of y test data generated from PrepareData function

  • time: numpy array of time test data generated from PrepareData function

  • xname: name of x data (defaults to X Data)

  • yname: name of y data to appear on plots (defaults to Y Data)

  • seqout: boolean to specify if y data is single values or sequences. If True, the last value in each sequence is used for plotting and metrics (defaults to False)

  • showplot: boolean to specify if the plots should be displayed or only saved (defaults to True)

Output:

  • Two plots are saved to the savepath showing predicted vs true value comparision

  • A text file with error metrics is saved to savepath