Reference

snoscience.metrics

This module contains functions to calculate network performance.

calculate_accuracy

Calculate the accuracy percentage from the given matrices.

Parameters:

Name	Type	Description	Default
calc	ndarray	Array containing all the calculated or predicted values.	required
true	ndarray	Array containing all the true or expected values.	required

Returns:

Name	Type	Description
accuracy	float	Accuracy percentage.

calculate_mse

Calculate the mean squared error per output vector from the given matrices.

Parameters:

Name	Type	Description	Default
calc	ndarray	Array containing all the calculated or predicted values.	required
true	ndarray	Array containing all the true or expected values.	required

Returns:

Type	Description
ndarray	Mean squared error per output.

snoscience.networks

This module contains the implementations of the neural network interfaces.

SimpleNeuralNetwork

Bases: NeuralNetwork

Single-process sequential feed-forward neural network.

__init__

Parameters:

Name	Type	Description	Default
loss	str	Loss function to use when training the network.	'mse'
optimiser	str	Optimiser to use when training the network.	'sgd'

Raises:

Type	Description
ValueError	Loss function is not supported.
ValueError	Optimiser is not supported.

predict

Let the neural network predict the outputs from the given inputs.

Parameters:

Name	Type	Description	Default
x	ndarray	Inputs for the network.	required
classify	bool	Create classification from output layer, otherwise keep regression.	required

Returns:

Type	Description
ndarray	Predictions from the network.

Raises:

Type	Description
ValueError	Size of the input array does not match with the network inputs.
ValueError	No layers are present in the network.
ValueError	Network is not trained.

train

Train the neural network with the given parameters.

Parameters:

Name	Type	Description	Default
x	ndarray	Input samples to train the network with.	required
y	ndarray	Output samples to train the network with.	required
epochs	int	Number of training iterations.	required
samples	int	Number of samples taken from the dataset per iteration.	required
kwargs		Optimiser hyperparameters as keyword arguments.	{}

Raises:

Type	Description
ValueError	Number of samples in the input array do not match with the output array.
ValueError	No layers are present in the network.
ValueError	Outputs per sample do not match with the number of neurons in the output layer.
ValueError	Number of samples per epoch is larger than the dataset.
KeyError	Hyperparameters required for optimiser are not given as kwargs.