Example #1
0
def main():
    # Parse arguments
    parser = ArgumentParser()
    parser.add_argument(
        '-cur_state_pkl',
        required=True,
        help='Path to the pickle file holding the current state.')
    parser.add_argument(
        '-prev_state_pkl',
        required=True,
        help='Path to the pickle file holding the previous state.')
    parser.add_argument('-global_step_db',
                        required=True,
                        help='Path to db holding global step results.')
    args, unknown = parser.parse_known_args()
    # raise ValueError(args)
    fname_cur_state = path.abspath(args.cur_state_pkl)
    fname_prev_state = path.abspath(args.prev_state_pkl)
    global_db = path.abspath(args.global_step_db)

    # Load local state
    local_state = StateData.load(fname_prev_state).data
    # Load global node output
    global_out = LogRegrIter_Glob2Loc_TD.load(global_db)
    # Run algorithm local iteration step
    local_state, local_out = logregr_local_iter(local_state=local_state,
                                                local_in=global_out)
    # Save local state
    local_state.save(fname=fname_cur_state)
    # Return
    local_out.transfer()
Example #2
0
def logregr_global_iter(global_state, global_in):
    # Unpack global state
    n_obs = global_state['n_obs']
    n_cols = global_state['n_cols']
    ll_old = global_state['ll']
    coeff = global_state['coeff']
    iter = global_state['iter']
    y_val_dict = global_state['y_val_dict']
    schema_X = global_state['schema_X']
    schema_Y = global_state['schema_Y']
    # Unpack global input
    ll_new, grad, hess = global_in.get_data()

    # Compute new coefficients
    coeff = np.dot(np.linalg.inv(hess), grad)
    # Update termination quantities
    delta = abs(ll_new - ll_old)
    iter += 1

    # Pack state and results
    global_state = StateData(n_obs=n_obs,
                             n_cols=n_cols,
                             ll=ll_new,
                             coeff=coeff,
                             delta=delta,
                             iter=iter,
                             y_val_dict=y_val_dict,
                             schema_X=schema_X,
                             schema_Y=schema_Y)
    global_out = LogRegrIter_Glob2Loc_TD(coeff)
    return global_state, global_out
Example #3
0
def logregr_global_init(global_in):
    n_obs, n_cols, y_val_dict, schema_X, schema_Y = global_in.get_data()

    if n_obs == 0:
        raise ExaremeError('The selected variables contain 0 datapoints.')

    # Init vars
    ll = - 2 * n_obs * np.log(2)
    coeff = np.zeros(n_cols)
    iter = 0

    # Pack state and results
    global_state = StateData(n_obs=n_obs, n_cols=n_cols, ll=ll, coeff=coeff, iter=iter,
                             y_val_dict=y_val_dict, schema_X=schema_X, schema_Y=schema_Y)
    global_out = LogRegrIter_Glob2Loc_TD(coeff)

    return global_state, global_out