NN to solve sin function

Author: edwardyang12

Summer20/NeuralNetwork/tfsin.py

@@ -0,0 +1,61 @@
+from tensorflow.keras import datasets, models, layers, losses
+import tensorflow as tf
+import math
+import numpy as np
+import random
+import matplotlib.pyplot as plt
+
+number = 100
+
+# randomPoints on the sin graph
+# number - number of points
+# angles - tuple for range of angles
+def randomPoints(number, angles= (-360,360)):
+    angleList = []
+    sinList = []
+    while(number>0):
+        angle = random.uniform(angles[0],angles[1])
+        angle = angle * math.pi / 180 # convert to radians
+        angleList.append([angle])
+        sinList.append([math.sin(angle)])
+        number = number -1
+    return angleList, sinList
+
+# only specific points on the sin graph
+# number - number of points
+# angles - target angle
+def selectPoints(number, angle=30):
+        angleList = []
+        sinList = []
+        while(number>0):
+            radangle = angle * math.pi / 180 # convert to radians
+            angleList.append([radangle])
+            sinList.append([math.sin(radangle)])
+            angleList.append([-radangle])
+            sinList.append([math.sin(-radangle)])
+            angle = angle + 60
+            number = number -1
+        return angleList, sinList
+
+angles = (-360,360)
+angleList, sinList = randomPoints(10000, angles)
+#angleList, sinList = selectPoints(10000)
+
+model = models.Sequential()
+model.add(layers.Dense(10, activation='tanh', input_shape=(1,)))
+model.add(layers.Dense(1, activation=None))
+model.compile(optimizer='Adam',
+                loss=losses.MeanSquaredError(),
+                metrics=['mean_squared_error'])
+history = model.fit(np.array(angleList),np.array(sinList), epochs=200)
+plt.plot(history.history['mean_squared_error'], label='mean_squared_error')
+plt.xlabel('Epoch')
+plt.ylabel('MSE')
+plt.ylim([0.0, 0.2])
+plt.legend(loc='lower right')
+
+angleTest, sinTest = randomPoints(10)
+print(model.predict(np.array(angleTest)))
+print(sinTest)
+
+plt.show()