I’m coding a neural network to add and subtract numbers. Below is the code:

import tensorflow as tf
from keras.models import Sequential
from keras.layers import Dense
import numpy as np

num_train = 1000

X_train = np.random.rand(num_train, 2)
y_train_add = X_train[:, 0] + X_train[:, 1]
y_train_subtract = X_train[:, 0] - X_train[:, 1]

model_add = Sequential(
        [
            Dense(10, activation='relu'),
            Dense(1, activation ='relu')
            ]
        )

model_subtract = Sequential(
        [
            Dense(10, activation='relu'),
            Dense(1, activation ='relu')
            ]
        )

batch_size = 32
epochs = 100

model_add.compile(loss = 'mse', optimizer='adam')
model_add.fit(X_train, y_train_add, batch_size=batch_size, epochs=epochs, verbose = 1)


model_subtract.compile(loss = 'mse', optimizer='adam')
model_subtract.fit(X_train, y_train_subtract, batch_size=batch_size, epochs=epochs, verbose = 1)

x = "n" 
while True:
    print("enter first num:")
    x = input()
    if x == "end":
        break

    print("Enter operation:")
    op= input()

    print("enter second num:")
    y = input()

    X = int(x)
    Y = int(y)

    if op == "+":
        predicted_sum = model_add.predict(np.array([[X, Y]]))
        print(predicted_sum)
    elif op =="-":
        predicted_sum = model_subtract.predict(np.array([[X, Y]]))
        print(predicted_sum)

There are two problems with it:

1: It works sometimes, but not always. On some occasions, when I run the program, it outputs approximately correct values for every pair of numbers I input. In other trials, every pair of numbers and operations gives [[0.]] as an output. There seems to be no pattern to it; sometimes when I run it, it just decides to output [[0.]], and it works fine at other times. I don’t even change the code between trials.

2: Every time I subtract two numbers with the result being negative, it outputs a small positive number.

Can someone help me fix this?