Implementing a simple RNN

We start with a simple form of a recurrent neural network to understand the basic idea of RNNs. In this example, we will feed the RNN four binary variables. These represent the weather types on a certain day. For example, [1, 0, 0, 0] stands for sunny and [1, 0, 1, 0] stands for sunny and windy. The target value is a double representing the percentage of rain on that day. For this problem, we can say that the quantity of rain on a certain day also depends on the values of the previous day. This makes this problem well suited for a 4-to-1 RNN model.

How to do it...

1. In this basic example, we will our simple RNN with NumPy:

import numpy as np

2. Let's start with creating the dummy dataset that we will be using:

X = []
X.append([1,0,0,0])
X.append([0,1,0,0])
X.append([0,0,1,0])
X.append([0,0,0,1])
X.append([0,0,0,1])
X.append([1,0,0,0])
X.append([0,1,0,0])
X.append([0,0,1,0])
X.append([0,0,0,1])

y = [0.20, 0.30, 0.40, 0.50, 0.05, 0.10, 0.20,
0.30, 0.40]

3. For this regression...