Activity 2.01: Build a Multilayer Neural Network to Classify Sonar Signals

In this activity, we will use the Sonar dataset (https://archive.ics.uci.edu/ml/datasets/Connectionist+Bench+(Sonar,+Mines+vs.+Rocks)), which has patterns obtained by bouncing sonar signals off a metal cylinder at various angles and under various conditions. You will build a neural network-based classifier to classify between sonar signals bounced off a metal cylinder (the Mine class), and those bounced off a roughly cylindrical rock (the Rock class). We recommend using the Keras API to make your code more readable and modular, which will allow you to experiment with different parameters easily:

Note

You can download the sonar dataset from this link https://packt.live/31Xtm9M.

1. The first step is to understand the data so that you can figure out whether this is a binary classification problem or a multiclass classification problem.
2. Once you understand the data and the type of classification that needs to be done, the next step is network configuration: the number of neurons, the number of hidden layers, which activation function to use, and so on.

   Recall the network configuration steps that we've covered so far. Let's just reiterate a crucial point, the activation function part: for the output (the last) layer, we use sigmoid to do binary classification and Softmax to do multiclass classification.

3. Open the sonar.csv file to explore the dataset and see what the target variables are.
4. Separate the input features and the target variables.
5. Preprocess the data to make it neural network-compatible. Hint: one-hot encoding.
6. Define a neural network using Keras and compile it with the right loss function.
7. Print out a model summary to verify the network parameters and considerations.

You are expected to get an accuracy value above 95% by designing a proper multilayer neural network using these steps.

Note

The detailed steps for this activity, along with the solutions and additional commentary, are presented on page 390.