kraken/backend/training.py

import logging
import joblib
import os
from datetime import datetime
from functools import partial
import yaml

import mne
import numpy as np
from scipy import signal
from pyriemann.estimation import BlockCovariances

import bci_core.feature_extractors as feature_extractors
import bci_core.utils as bci_utils
import bci_core.model as bci_model
from dataloaders import neo


def train_model(raw, event_id, model_type='baseline'):
    """
    """
    events, _ = mne.events_from_annotations(raw, event_id=event_id)
    if model_type.lower() == 'baseline':
        model = _train_baseline_model(raw, events)
    elif model_type.lower() == 'riemann':
        # TODO: load subject config
        model = _train_riemann_model(raw, events)
    else:
        raise NotImplementedError
    return model


def _train_riemann_model(raw, events, lfb_bands=[(15, 30), [30, 45]], hg_bands=[(55, 95), (105, 145)]):
    fs = raw.info['sfreq']
    n_ch = len(raw.ch_names)
    feat_extractor = feature_extractors.FeatExtractor(fs, lfb_bands, hg_bands)
    filtered_data = feat_extractor.transform(raw.get_data())
    # TODO: find proper latency
    X = bci_utils.cut_epochs((0, 1., fs), filtered_data, events[:, 0])
    y = events[:, -1]
    
    scaler = bci_model.ChannelScaler()
    X = scaler.fit_transform(X)

    # compute covariance
    lfb_dim = len(lfb_bands) * n_ch
    hgs_dim = len(hg_bands) * n_ch
    cov_model = BlockCovariances([lfb_dim, hgs_dim], estimator='lwf')
    X_cov = cov_model.fit_transform(X)

    param = {'C_lfb': np.logspace(-4, 0, 5), 'C_hgs': np.logspace(-3, 1, 5)}

    model_func = partial(bci_model.stacking_riemann, lfb_dim=lfb_dim, hgs_dim=hgs_dim)
    best_auc, best_param = bci_utils.param_search(model_func, X_cov, y, param)

    logging.info(f'Best parameter: {best_param}, best auc {best_auc}')

    # train and dump best model
    model_to_train = model_func(**best_param)
    model_to_train.fit(X, y)
    return [feat_extractor, scaler, cov_model, model_to_train]


def _train_baseline_model(raw, events):
    fs = raw.info['sfreq']
    filter_bank_data = feature_extractors.filterbank_extractor(raw.get_data(), fs, np.arange(20, 150, 15), reshape_freqs_dim=True)

    filter_bank_epoch = bci_utils.cut_epochs((0, 1., fs), filter_bank_data, events[:, 0])

    # downsampling to 10 Hz
    # decim 2 times, to 100Hz
    decimate_rate = np.sqrt(fs / 10).astype(np.int16)
    filter_bank_epoch = signal.decimate(filter_bank_epoch, decimate_rate, axis=-1, zero_phase=True)
    # to 10Hz
    filter_bank_epoch = signal.decimate(filter_bank_epoch, decimate_rate, axis=-1, zero_phase=True)
    X = filter_bank_epoch
    y = events[:, -1]

    best_auc, best_param = bci_utils.param_search(bci_model.baseline_model, X, y, {'C': np.logspace(-5, 4, 10)})
    logging.info(f'Best parameter: {best_param}, best auc {best_auc}')

    model_to_train = bci_model.baseline_model(**best_param)
    model_to_train.fit(X, y)
    return model_to_train


def model_saver(model, model_path, model_type, subject_id, event_id):
    # event list should be sorted by class label
    sorted_events = sorted(event_id.items(), key=lambda item: item[1])
    # Extract the keys in the sorted order and store them in a list
    sorted_events = [item[0] for item in sorted_events]

    try:
        os.mkdir(os.path.join(model_path, subject_id))
    except FileExistsError:
        pass

    now = datetime.now()
    classes = '+'.join(sorted_events)
    date_time_str = now.strftime("%m-%d-%Y-%H-%M-%S")
    model_name = f'{model_type}_{classes}_{date_time_str}.pkl'
    joblib.dump(model, os.path.join(model_path, subject_id, model_name))


if __name__ == '__main__':
    # TODO: argparse
    subj_name = 'ylj'
    model_type = 'baseline'
    # TODO: load subject config

    data_dir = f'./data/{subj_name}/train/'
    model_dir = './static/models/'

    info = yaml.safe_load(os.path.join(data_dir, 'info.yml'))
    sessions = info['sessions']
    event_id = {'rest': 0}
    for f in sessions.keys():
        event_id[f] = neo.FINGERMODEL_IDS[f]
    
    # preprocess raw
    raw = neo.raw_preprocessing(data_dir, sessions, rename_event=False, ori_epoch_length=5)

    # train model
    model = train_model(raw, event_id=event_id, model_type=model_type)
    
    # save
    model_saver(model, model_dir, model_type, subj_name, event_id)