LoginSignup
0

More than 5 years have passed since last update.

kerasでbackend その2

Last updated at Posted at 2018-01-02

概要

kerasのbackendだけで、学習して分類してみた。

写真

backend11.png

環境

Raspberry Pi 3 model B v1.2 element14
2017-09-07-raspbian-stretch
tensorflow-1.3

サンプルコード

from tensorflow.contrib.keras.python.keras import backend as K
from tensorflow.contrib.keras.python.keras.optimizers import SGD
import numpy as np
import matplotlib.pyplot as plt

dx, dy = [], []
dx.append([0.0, 0.0])
dx.append([0.0, 1.0])
dx.append([1.0, 0.0])
dx.append([1.0, 1.0])
dy.append(0.0)
dy.append(1.0)
dy.append(1.0)
dy.append(0.0)
input_dim = 2
output_dim = 1
hidden_dim = 8
x = K.placeholder(shape = (None, input_dim), name = "x")
ytrue = K.placeholder(shape = (None, output_dim), name = "y")
W1 = K.random_uniform_variable((input_dim, hidden_dim), 0, 1, name = "W1")
W2 = K.random_uniform_variable((hidden_dim, output_dim), 0, 1, name = "W2")
b1 = K.random_uniform_variable((hidden_dim, ), 0, 1, name = "b1")
b2 = K.random_uniform_variable((output_dim, ), 0, 1, name = "b2")
params = [W1, b1, W2, b2]
hidden = K.tanh(K.dot(x, W1) + b1)
ypred = K.tanh(K.dot(hidden, W2) + b2)
loss = K.mean(K.square(ypred - ytrue), axis = -1)
opt = SGD()
updates = opt.get_updates(params, [], loss)
train = K.function(inputs = [x, ytrue], outputs = [loss], updates = updates)
for ep in range(1000):
    for i in range(4):
        c3 = train([[dx[i]], [[dy[i]]]])
    if ep % 100 == 0:
        print (ep, c3[0])
pred = K.function(inputs = [x], outputs = [ypred])

sample = 4
xps, yps = [], []
for i in range(sample):
    if dy[i] < 0.5:
        xps.append(dx[i][0])
        yps.append(dx[i][1])
xns, yns = [], []
for i in range(sample):
    if dy[i] >= 0.5:
        xns.append(dx[i][0])
        yns.append(dx[i][1])
ticks = 40
a = []
b = []
for ix in range(ticks):
    for iy in range(ticks):
        po = [ix * 2.0 / ticks - 0.5, iy * 2.0 / ticks - 0.5]
        a.append(po)
        preds = pred([[po]])
        b.append(preds)
p = np.array(a)
r = np.array(b)
thresh = 0.5
xp, yp = [], []
for i in range(ticks ** 2):
    if r[i]< thresh:
        xp.append(p[i][0])
        yp.append(p[i][1])
plt.scatter(xp, yp, c = "cyan", marker = ".")
xn, yn = [], []
for i in range(ticks ** 2):
    if r[i] >= thresh:
        xn.append(p[i][0])
        yn.append(p[i][1])
plt.scatter(xn, yn, c = "magenta", marker = ".")
plt.scatter(xps, yps, c = "blue", marker = "o", s = 100)
plt.scatter(xns, yns, c = "red", marker = "o", s = 200)
plt.savefig("backend11.png")
plt.show()



以上。

Register as a new user and use Qiita more conveniently

  1. You get articles that match your needs
  2. You can efficiently read back useful information
  3. You can use dark theme
What you can do with signing up
0