Module project.test.esn_t

import itertools as it
from math import inf
from pprint import pprint

import numpy as np
from matplotlib.pyplot import *
from scipy import linalg, sparse

import project.esn.core as c
import project.esn.matrix as m
import project.esn.trainer as tr
import project.esn.transformer as ta
import project.esn.updater as up
import project.music_gen.core as cgen
import project.parse_midi.matrix.core as cmidi
import project.parse_midi.matrix.proc_dicts as emidi
import project.test.music_test as tgen


def multiple__(times=10, thr=lambda x: True):
    def decorator(fun):
        def inner():
            filt = [x for x in [fun()[1] for _ in range(times)] if thr(x)]
            return (times, len(filt), np.mean(filt), filt)

        return inner

    return decorator


@multiple__(thr=10)
def test():
    data = np.loadtxt('/home/vimmoos/NN/testESN/MackeyGlass_t17.txt')
    with c.Run(
            **{
                "data": data,
                "in_out": 1,
                "reservoir": 100,
                "train_len": 2000,
                "test_len": 2000,
                "init_len": 100,
                "leaking_rate": 0.3,
                "spectral_radius": 1.25,
                "density": .5,
                "reg": 1e-8
            }) as gen:
        return gen()


@multiple__()
def test_randomMatrix():
    import random
    train_len = test_len = 2000
    data = np.array([
        np.array([random.randint(0, 10) / 10])
        for _ in range(train_len + test_len + 1)
    ])
    with c.Run(
            **{
                "data": data,
                "in_out": 1,
                "reservoir": 100,
                "train_len": train_len,
                "test_len": test_len,
                "init_len": 100,
                "leaking_rate": 0.3,
                "spectral_radius": 1.25,
                "density": .5,
                "reg": 1e-8
            }) as gen:
        return gen()


def test_midi():
    train_len = test_len = 970
    init_len = 100
    music = it.repeat(
        cmidi.exec_proc_dict(emidi.example_proc_dict)["matrixs"][0], 20)
    data = c.Data(np.array(list(music)), None, init_len, train_len, test_len)
    with c.Run(
            **{
                "data": data,
                "in_out": 6,
                "reservoir": 500,
                "error_len": 500,
                "leaking_rate": 0.3,
                "spectral_radius": 0.8,
                "density": .5,
                "reg": 1e-8,
                "transformer": ta.user_threshold(0.5),
            }) as gen:
        return gen()


# @multiple__(thr=lambda x : sum(x) > 4)
def test_generated():
    train_len = test_len = 1200
    init_len = 200
    music = (tgen.test_patterns[2] * 300)

    data = c.Data(np.array(list(~music)), music.tempo, init_len, train_len,
                  test_len)
    with c.Run(
            **
        {
            "data": data,
            "reservoir": 300,
            "in_out": 9,
            "leaking_rate": 0.3,
            "reg": 1e-8,
            "transformer": ta.Transformers.pow_prob,
            "t_param": 1,
            "t_squeeze": np.tanh,
        }).load(
            "/home/vimmoos/NN/resources/reservoir/0.18333333333333335_0.04_2000",
            0) as gen:
        return gen()


# tot, ind = run_multiple(10, test_generated)