Hands-On Genetic Algorithms with Python

Solution representation and evaluation

Since we have decided to control the cart in the CartPole challenge using a neural network of the Multilayer Perceptron type, the set of parameters that we will need to optimize are the network's weights and biases, as follows:

• Input layer: This layer does not participate in the network mapping; instead, it receives the input values and forwards them to every neuron in the next layer. Therefore, no parameters are needed for this network.
• Hidden layer: Each node in this layer is fully connected to each of the inputs, and therefore requires four weights in addition to a single bias value.
• Output layer: The single node in this layer is connected to each of the nodes in the hidden layer, and therefore requires ...