这是我通过跟踪sentdex视频实现的神经网络，然后使用我自己的数据集进行修改以进行训练。它的问题是当我为我的csv数据集运行这个网络时，它要么给出精度1.0或0.0。
代码
from __future__ import print_function
from python_speech_features import mfcc
from python_speech_features import delta
from python_speech_features import logfbank
import scipy.io.wavfile as wav
import numpy as np
import matplotlib.pyplot as pt
import csv
import csv
from sklearn.preprocessing import normalize
import tensorflow as tf
# Parameters
learning_rate = 0.0001
training_epochs = 30
batch_size = 100
display_step = 1
# Network Parameters
n_nodes_hl1 = 100
n_nodes_hl2 = 100
n_nodes_hl3 = 100 # 2nd layer number of features
n_input = 13 # MNIST data input (img shape: 28*28)
n_classes = 2 # MNIST total classes (0-9 digits)
a = []
f = open('C:\\Users\\Monil\\Desktop\\python_speech_features-master\\LatestFeatureSet.csv', 'r')
try:
reader = csv.reader(f)
for row in reader:
a.append(list(row))
finally:
f.close()
print(len(a))
X = [[0.0] * 13] * len(a)
for i in range(len(a)):
for j in range(13):
X[i][j] = float(a[i][j])
Y=[]
for i in range(len(a)):
Y.append(a[i][-1])
data = normalize(X)
target = [[0] * 2] * len(a)
for i in range(len(a)):
if Y[i] == 0:
target[i][0] = 1
target[i][1] = 0
else:
target[i][0] = 0
target[i][1] = 1
print(data)
print(target)
x = tf.placeholder("float", [None, n_input])
y = tf.placeholder("float", [None, n_classes])
# Create model
def multilayer_perceptron(x,weights,biases):
layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])
layer_1 = tf.nn.relu(layer_1)
# Hidden layer with RELU activation
layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])
layer_2 = tf.nn.relu(layer_2)
layer_3 = tf.add(tf.matmul(layer_2, weights['h3']), biases['b3'])
layer_3 = tf.nn.relu(layer_3)
# Output layer with linear activation
out_layer = tf.matmul(layer_3, weights['out']) + biases['out']
return out_layer
# Store layers weight & bias
weights = {
'h1': tf.Variable(tf.random_normal([n_input, n_nodes_hl1])),
'h2': tf.Variable(tf.random_normal([n_nodes_hl1, n_nodes_hl2])),
'h3': tf.Variable(tf.random_normal([n_nodes_hl2, n_nodes_hl3])),
'out': tf.Variable(tf.random_normal([n_nodes_hl3, n_classes]))
}
biases = {
'b1': tf.Variable(tf.random_normal([n_nodes_hl1])),
'b2': tf.Variable(tf.random_normal([n_nodes_hl2])),
'b3': tf.Variable(tf.random_normal([n_nodes_hl3])),
'out': tf.Variable(tf.random_normal([n_classes]))
}
# Construct model
pred = multilayer_perceptron(x,weights,biases)
# Define loss and optimizer
cost = tf.reduce_mean(tf.squared_difference(y, pred))
optimizer = tf.train.AdadeltaOptimizer(learning_rate).minimize(cost)
# Initializing the variables
init = tf.global_variables_initializer()
# Launch the graph
with tf.Session() as sess:
sess.run(init)
# Training cycle
for epoch in range(training_epochs):
avg_cost = 0.
total_batch = int(len(data) / batch_size)
# Loop over all batches
for i in range(total_batch):
batch_x, batch_y = data[batch_size * i:batch_size * (i + 1)], target[batch_size * i:batch_size * (i + 1)]
# Run optimization op (backprop) and cost op (to get loss value)
_, c = sess.run([optimizer, cost], feed_dict={x: batch_x,
y: batch_y})
# Compute average loss
#print("c is : ",c)
avg_cost += c / (total_batch-1)
# Display logs per epoch step
if epoch % display_step == 0:
print("Epoch:", '%04d' % (epoch + 1), "cost=", \
"{:.4f}".format(avg_cost))
print("Optimization Finished!")
f = open('C:\\Users\\Monil\\Desktop\\python_speech_features-master\\LatestFeatureTestSet.csv', 'r')
try:
reader = csv.reader(f)
for row in reader:
a.append(list(row))
finally:
f.close()
print(len(a))
X = [[0.0] * 13] * len(a)
for i in range(len(a)):
for j in range(13):
X[i][j] = float(a[i][j])
Y = []
for i in range(len(a)):
Y.append(a[i][-1])
data = normalize(X)
target = [[0] * 2] * len(a)
for i in range(len(a)):
if Y[i] == 0:
target[i][0] = 1
target[i][1] = 0
else:
target[i][0] = 0
target[i][1] = 1
#print(data)
#print(target)
# Test model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
print(correct_prediction)
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
print("{:.4f}".format(accuracy.eval(feed_dict={x: data, y:target})))
sess.close()
培训输出：
Epoch: 0001 cost= 44004.7984
Epoch: 0002 cost= 41676.6541
Epoch: 0003 cost= 38812.2610
Epoch: 0004 cost= 35557.2471
Epoch: 0005 cost= 31927.6941
Epoch: 0006 cost= 28231.0714
Epoch: 0007 cost= 24604.1943
Epoch: 0008 cost= 21078.5401
Epoch: 0009 cost= 17752.3665
Epoch: 0010 cost= 14660.3286
Epoch: 0011 cost= 11828.9188
Epoch: 0012 cost= 9282.4267
Epoch: 0013 cost= 7046.8689
Epoch: 0014 cost= 5154.4891
Epoch: 0015 cost= 3584.8977
Epoch: 0016 cost= 2341.5776
Epoch: 0017 cost= 1435.0431
Epoch: 0018 cost= 792.7084
Epoch: 0019 cost= 378.7035
Epoch: 0020 cost= 139.6612
Epoch: 0021 cost= 31.0557
Epoch: 0022 cost= 2.2325
Epoch: 0023 cost= 0.0033
Epoch: 0024 cost= 0.0000
Epoch: 0025 cost= 0.0000
Epoch: 0026 cost= 0.0000
Epoch: 0027 cost= 0.0000
Epoch: 0028 cost= 0.0000
Epoch: 0029 cost= 0.0000
Epoch: 0030 cost= 0.0000
Optimization Finished!
220697
Tensor("Equal:0", shape=(?,), dtype=bool)
1.0000