Create a Simple Recurrent Neural Network using Kears

In this tutorial, we will explore the architecture and inner workings of RNNs, understand the concept of recurrent layers, and build a basic RNN model using Keras.

What is RNN

RNN stands for Recurrent Neural Network. It combines the current input with the previous hidden state, updates the hidden state, and produces an output for each step. Advanced variants like LSTM and GRU have been developed to improve the RNN’s ability to capture long-term dependencies.

Import the Libraies

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

Model of RNN

model = Sequential()
model.add(SimpleRNN(units=32, input_shape=(10,3)))
model.add(Dense(units=1, activation='sigmoid'))
model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])

• Sequential() – This line initializes the sequential model. It provides an easy way to build a neural network by adding layers sequentially.
• model.add(SimpleRNN(units=32, input_shape=(10,3))) – Add layers to the model. units determine the number of neurons in simpleRNN layers.input_shape specifies the shape of the input data in the first layer.(10, 3) indicating that the input data has a sequence length of 10 and each step in the sequence has 3 features.
• model.add(Dense(units=1, activation='sigmoid')) – Dense layers added. Single neurons are present in the output layer. activation sigmoid means producing an output value between 0 and 1.
• model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy']) – The compile() function is used to configure the model for training. The loss parameter is set to ‘binary_crossentropy' which is the loss function used for binary classification tasks. The metrics parameter is set to ['accuracy']indicate that the model’s performance will be evaluated based on accuracy during training.

model.summary()

• model.summary() – Print the summary of the model.

Output