This repo contains:

• LGM_kinect.ipynb: Notebook with the python code functions of each task.
• data: data directory that contains:

  • data.mat: file containing the data for the project. It can be opened with function 'load_dataset' in notebook file. The data includes: position of joints (20 x 3 x 2045) corresponding to 20 joints x 3 positions (x,y,z) x 2045 individuals; labels 2045 and person index (2045 vector) indicating what is the person doing the action

  • validation_data.mat: Contains an example of NB model and LG model for the data included in the file. Useful to validate the learning implementation. It can load it as:

    import scipy.io

    dd = scipy.io.loadmat('data/validation_data.mat') dd['data_small'] # Input data dd['labels_small'] # Input labels dd['individuals_small'] # Input individual indexes dd['train_indexes'] # Instances used for training dd['test_indexes'] # Instances used for test dd['model_nb'] # NB model dd['model_lg'] # LG model dd['accur_nb'] # Accuracy of NB model on test instances dd['accur_lg'] # Accuracy of LG model on test instances

    import scipy.io

    dd = scipy.io.loadmat('data/ejemplolineargaussian.mat') dd = dd['ejemplo'][0]

    The inputs are

    X = dd['inputX'] Y = dd['inputY']

    The expected outputs are

    betas = dd['outputBetas'] sigma = dd['outputSigma']