__author__ = 'martslaaf' import numpy as np from matplotlib.pyplot import plot, show from wavenet import wavelon_class_constructor, trainer coun = 3000 inp_s = map(lambda x: float(x.split(',')[0]), open('/home/martslaaf/Learn_Coding/pybrain/sig.csv').readlines()) outp_s = map(lambda x: float(x.split(',')[1]), open('/home/martslaaf/Learn_Coding/pybrain/sig.csv').readlines()) inp_o = map(lambda x: float(x.split(',')[0])-150, open('/home/martslaaf/Learn_Coding/pybrain/orig.csv').readlines()) outp_o = map(lambda x: float(x.split(',')[1]), open('/home/martslaaf/Learn_Coding/pybrain/orig.csv').readlines()) ds = [] for i in xrange(coun): ds.append((np.array([inp_s[i]]), np.array([outp_s[i]]))) vs = [] for i in xrange(301): vs.append((np.array([inp_o[i]]), np.array([outp_o[i]]))) n = wavelon_class_constructor() n = n(1, 1, 19) k = 0 track = trainer(300, ds, vs, n) outew = [] print track for j in vs: outew.append(n.forward(j[0])[0][0]) plot(outp_o, 'g') plot(outew, 'r') show()
__author__ = "martslaaf" import numpy as np from matplotlib.pyplot import plot, show from wavenet import wavelon_class_constructor, trainer inp_1 = map(lambda x: float(x), open("/home/martslaaf/Pictures/old_data/nonlinear_xor_1.csv").readlines()) inp_2 = map(lambda x: float(x), open("/home/martslaaf/Pictures/old_data/nonlinear_xor_2.csv").readlines()) inp_3 = map(lambda x: float(x), open("/home/martslaaf/Pictures/old_data/nonlinear_sum.csv").readlines()) outp = map(lambda x: float(x), open("/home/martslaaf/Pictures/old_data/nonlinear_target.csv").readlines()) coun = 1000 tr = [] va = [] for i in xrange(coun - 250): tr.append((np.array([inp_1[i], inp_2[i], inp_3[i]]), np.array([outp[i]]))) for i in xrange(coun - 250, coun): va.append((np.array([inp_1[i], inp_2[i], inp_3[i]]), np.array([outp[i]]))) n = wavelon_class_constructor(frame=(-200, 200), period=100) n = n(3, 1, 19) k = 0 track = trainer(10000, tr, va, n) outew = [] print track for j in va: outew.append(n.forward(j[0])[0][0]) plot(outp[coun - 250 : coun], "g") plot(outew, "r") show()
vs = [] for i in xrange(301): vs.append((np.array([inp_o[i]]), np.array([outp_o[i]]))) seed() # networks initialization exp_1_n = [] # default set exp_2_n = [] # default but Mhat exp_3_n = [] # random limits exp_4_n = [] # hidden layer x2 exp_5_n = [] # right limits for translation exp_6_n = [] # period data (Nyqist) exp_7_n = [] # fourier analysis exp_8_n = [] # hidden layer /2 mini, maxi = -150, 150 e_1 = wavelon_class_constructor() e_2 = wavelon_class_constructor(motherfunction=Mhat) e_3 = wavelon_class_constructor(frame=(uniform(-100, 0), uniform(0, 100)), period=uniform(0, 100)) e_4 = wavelon_class_constructor() e_5 = wavelon_class_constructor(frame=(mini, maxi)) e_6 = wavelon_class_constructor(period=120) e_7 = wavelon_class_constructor(period=120, signal=outp_s, fa=True) e_8 = wavelon_class_constructor() seed() for i in xrange(repeat_count): exp_1_n.append(e_1(1, 1, 19)) exp_2_n.append(e_2(1, 1, 19)) exp_3_n.append(e_3(1, 1, 38)) exp_4_n.append(e_4(1, 1, 19)) exp_5_n.append(e_5(1, 1, 19)) exp_6_n.append(e_6(1, 1, 19))
from matplotlib.pyplot import plot, show from wavenet import wavelon_class_constructor, trainer coun = 3000 inp_s = map(lambda x: float(x.split(',')[0]), open('/home/martslaaf/Learn_Coding/pybrain/sig.csv').readlines()) outp_s = map(lambda x: float(x.split(',')[1]), open('/home/martslaaf/Learn_Coding/pybrain/sig.csv').readlines()) inp_o = map(lambda x: float(x.split(',')[0]) - 150, open('/home/martslaaf/Learn_Coding/pybrain/orig.csv').readlines()) outp_o = map(lambda x: float(x.split(',')[1]), open('/home/martslaaf/Learn_Coding/pybrain/orig.csv').readlines()) ds = [] for i in xrange(coun): ds.append((np.array([inp_s[i]]), np.array([outp_s[i]]))) vs = [] for i in xrange(301): vs.append((np.array([inp_o[i]]), np.array([outp_o[i]]))) n = wavelon_class_constructor() n = n(1, 1, 19) k = 0 track = trainer(300, ds, vs, n) outew = [] print track for j in vs: outew.append(n.forward(j[0])[0][0]) plot(outp_o, 'g') plot(outew, 'r') show()