utils package

Submodules

utils.ModelScaler module

class utils.ModelScaler.ModelScaler(xdim=None)

Bases: object

Normalise the inputs and outputs to the NN model.

fit(inputs, targets)

Fits the scaler to the model inputs and targets.

Parameters:

• inputs (np.array or tf.Tensor) – shape N x 16 + 2
• targets (np.array or tf.Tensor) – shape N x 16

get_vars()

Returns the tf.variables of the scaler objects used

inverse_transformInput(input)

Returns the inverse transform of the inputs

Parameters:input (np.array or tf.Tensor) – shape N x nS + nU or M x N x nS + nU Returns:shape N x nS + nU or M x N x nS + nU Return type:np.array or tf.Tensor

inverse_transformOutput(mean, variance)

Normalises the inverse transform of the targets to the NN model.

Parameters:

• mean (np.array or tf.Tensor) – shape N x nS
• variance (np.array or tf.Tensor) – shape N x nS

Returns:

shape N x nS

Return type:

np.array or tf.Tensor

transformInput(input)

Normalises the inputs to the NN model.

Parameters:input (np.array or tf.Tensor) – shape N x nS + nU or M x N x nS + nU Returns:shape N x nS + nU or M x N x nS + nU Return type:np.array or tf.Tensor

transformTarget(targets)

Normalises the targets to the NN model.

Parameters:targets (np.array or tf.Tensor) – shape N x nS Returns:shape N x nS Return type:np.array or tf.Tensor

utils.TensorStandardScaler module

class utils.TensorStandardScaler.TensorStandardScaler(x_dim, name=0)

Bases: object

Helper class for automatically normalizing inputs into the network.

cache()

Caches current values of this scaler.

Returns: None.

fit(data)

Runs two ops, one for assigning the mean of the data to the internal mean, and another for assigning the standard deviation of the data to the internal standard deviation. This function must be called within a ‘with <session>.as_default()’ block.

Arguments: data (np.ndarray): A numpy array containing the input

Returns: None.

get_vars()

Returns a list of variables managed by this object.

Returns: (list<tf.Variable>) The list of variables.

inverse_transform(data)

Undoes the transformation performed by this scaler.

Arguments: data (np.array): A numpy array containing the points to be transformed.

Returns: (np.array) The transformed dataset.

load_cache()

Loads values from the cache

Returns: None.

transform(data)

Transforms the input matrix data using the parameters of this scaler.

Arguments: data (np.array): A numpy array containing the points to be transformed.

Returns: (np.array) The transformed dataset.

class utils.TensorStandardScaler.TensorStandardScaler1D(name=0)

Bases: utils.TensorStandardScaler.TensorStandardScaler

Helper class for automatically normalizing inputs into the network.

fit(data)

Runs two ops, one for assigning the mean of the data to the internal mean, and another for assigning the standard deviation of the data to the internal standard deviation. This function must be called within a ‘with <session>.as_default()’ block.

Arguments: data (np.ndarray): A numpy array containing the input

Returns: None.

Module contents