heron package¶
Submodules¶
heron.acquisition module¶
heron.data module¶
The data module is designed to load and prepare arbitrary data sets for use in machine learning algorithms.

class
heron.data.
Data
(targets, labels, target_sigma=None, label_sigma=None, target_names=None, label_names=None, test_targets=None, test_labels=None, test_size=0.05)[source]¶ Bases:
object
The data class is designed to hold nontimeseries data, and is capable of automatically selecting test data from the provided dataset.
Future development will include the ability to add preselected test and verification data to the object.

add_data
(targets, labels, target_sigma=None, label_sigma=None)[source]¶ Add new rows into the data object.
 targetsarraylike
An array of training targets or “x” values which are to be used to train a machine learning algorithm.
 labelsarraylike
An array of training labels or “y” values which represent the observations made at the target locations of the data set.
 target_sigmaarraylike
Either an array of the uncertainty for each target point, or an array of the uncertainties, as a float, for each column in the targets.
 label_sigmaarraylike
Either an array of the uncertainty for each target point, or an array of the uncertainties, as a float, for each column in the labels.

calculate_normalisation
(data, name)[source]¶ Calculate the offsets for the normalisation. We’ll normally want to normalise the training data, and then be able to normalise and denormalise new inputs according to that.
 dataarraylike
The array of data to use to calculate the normalisations.
 namestr
The name to label the constants with.

denormalise
(data, name)[source]¶ Reverse the normalise() method’s effect on the data, and return it to the correct scaling.
 dataarraylike
The normalised data
 scalearraylike
The scalefactors used to normalise the data.
 arraylike
The denormalised data

get_starting
()[source]¶ Attempts to guess sensible starting values for the hyperparameter values.
 hyperparametersndarray
An array of values for the various hyperparameters.

normalise
(data, name)[source]¶ Normalise a given array of data so that the values of the data have a minimum at 0 and a maximum at 1. This improves the computability of the majority of data sets.
 dataarraylike
The array of data to be normalised.
 namestr
The name of the normalisation to be applied, e.g. training or label
 norm_dataarraylike
An array of normalised data.
 scale_factorsarraylike
An array of scale factors. The first is the DC offset, while the second is the multiplicative factor.
In order to perform the normalisation we need two steps: 1) Subtract the “DC Offset”, which is the minimum of the data 2) Divide by the range of the data


class
heron.data.
Timeseries
(targets, labels, target_names=None, label_names=None, test_size=0.05)[source]¶ Bases:
object
This is a class designed to hold timeseries data for machine learning algorithms.
Timeseries data needs to be handled differently from other datasets as it is rarely likely to be advantageous to select individual points from a timeseries as either test data or verification data. Instead the timeseries class will select individual timeseries as the test and verification data.
heron.kernels module¶

class
heron.kernels.
ExponentialSineSq
(period=1, width=15, ax=0)[source]¶ Bases:
heron.kernels.Kernel
An implementation of the exponential sinesquared kernel.

hyper
= [1, 1]¶

matrix
(data1, data2)[source]¶ Produce a gram matrix based off this kernel.
 datandarray
An array of data (x)
 covarndarray
A covariance matrix.

name
= 'Exponential sinesquared kernel'¶


class
heron.kernels.
Kernel
[source]¶ Bases:
object
A generic factory for Kernel classes.

distance
(data1, data2, hypers=None)[source]¶ Calculate the squared distance to the point in parameter space.

matrix
(data1, data2)[source]¶ Produce a gram matrix based off this kernel.
 datandarray
An array of data (x)
 covarndarray
A covariance matrix.

name
= 'Generic kernel'¶

ndim
= 1¶


class
heron.kernels.
Matern
(order=1.5, amplitude=100, width=15)[source]¶ Bases:
heron.kernels.Kernel
An implementation of the Matern Kernel.

name
= 'Matern'¶

order
= 1.5¶

heron.priors module¶

class
heron.priors.
Normal
(mean, std)[source]¶ Bases:
heron.priors.Prior
A normal prior probability distribution.
heron.regression module¶

class
heron.regression.
MultiTaskGP
(training_data, kernel, tikh=1e06, solver=<class 'george.solvers.hodlr.HODLRSolver'>, hyperpriors=None)[source]¶ Bases:
heron.regression.SingleTaskGP
An implementation of a cotrained set of Gaussian processes which share the same hyperparameters, but which model differing data. The training of these models is described in RW pp115–116.
A multitask GPR is capable of acting as a surrogate to a manytomany function, and is trained by making the assumption that all of the outputs from the function share a common correlation structure.
The principle difference compared to a single task GP is the presence of multiple Gaussian Processes, with one to model each dimension of the output data.
The MultiTask GPR implementation is very much a work in progress at the moment, and not all methods implemented in the SingleTask GPR are implemented correctly yet.

get_hyperparameters
()[source]¶ Return the kernel hyperparameters. Returns the hyperparameters of only the first GP in the network; the others /should/ all be the same, but there might be something to be said for checking this.
 hyperslist
A list of the kernel hyperparameters

ln_likelihood
(p)[source]¶ Provides a wrapper to the ln_likelihood functions for each component Gaussian process in the multitask system.
This is implemented in a separate function because of the mild peculiarities of how the pickle module needs to serialise functions, which means that instancemethods (which this would become) can’t be serialised.

prediction
(new_datum)[source]¶ Produce a prediction at a new point, or set of points.
 new_datumarray
The coordinates of the new point(s) at which the GPR model should be evaluated.
 prediction meansarray
The mean values of the function drawn from the Gaussian Process.
 prediction variancesarray
The variance values for the function drawn from the GP.

set_hyperparameters
(hypers)[source]¶ Set the hyperparameters of the kernel function on each Gaussian process.

train
(method='MCMC', metric='loglikelihood', sampler='ensemble', **kwargs)[source]¶ Train the Gaussian process by finding the optimal values for the kernel hyperparameters.
 methodstr {“MCMC”, “MAP”}
The method to be employed to calculate the hyperparameters.
 metricstr
The metric which should be used to assess the model.
 hyperpriorslist
The hyperprior distributions for the hyperparameters. Defaults to None, in which case the prior is uniform over all real numbers.


class
heron.regression.
Regressor
(training_data, kernel, tikh=1e06, solver=<class 'george.solvers.hodlr.HODLRSolver'>, hyperpriors=None)[source]¶

class
heron.regression.
SingleTaskGP
(training_data, kernel, tikh=1e06, solver=<class 'george.solvers.hodlr.HODLRSolver'>, hyperpriors=None)[source]¶ Bases:
object
This is an implementaion of a Single task Gaussian process regressor. That is, a GPR which is capable of acting as a surrogate to a manytoone function. The Single Task GPR is the fundamental building block of the MultiTask GPR, which consists of multiple Single Tasks which are trained in tandem (but which do NOT share correlation information). — Ahem… There /are/ components of this code in here, but things need a little bit more thought before this will work efficiently… An implementation of a Gaussian Process Regressor with multiple response outputs and multiple inputs.

active_learn
(afunction, x, y, iters=1, afunc_args={})[source]¶ Actively train the Gaussian process from a set of provided labels and targets using some acquisition function.
 afunctionfunction
The acquisition function.
 xarraylike
The input labels of the data. This can be a multidimensional array.
 yarraylike
The input targets of the data. This can only be a singledimensional array at present.
 itersint
The number of times to iterate the learning process: equivalently, the number of training points to digest.
 afunc_argsdict
A dictionary of arguments for the acquisition function. Optional.

correlation
()[source]¶ Calculate the correlation between the model and the test data.
 corrfloat
The correlation squared.

entropy
()[source]¶ Return the entropy of the Gaussian Process distribution. This can be calculated directly from the covariance matrix, making this a nice, quick calculation to perform.
 entropyfloat
The differential entropy of the GP.

expected_improvement
(x)[source]¶ Returns the expected improvement at the design vector X in the model
 xarraylike
A real world coordinates design vector
 EI: float
The expected improvement value at the point x in the model

grad_neg_ln_likelihood
(p)[source]¶ Return the negative of the gradient of the log likelihood for the GP when its hyperparameters have some specified value.
 gpheron Regressor object
The gaussian process to be evaluated
 parraylike
An array of the hyperparameters at which the model is to be evaluated.
 grad_ln_likelihoodfloat
The gradient of loglikelihood for the Gaussian process

hyperpriortransform
(p)[source]¶ Return the true value in the desired hyperprior space, given an input of a unithypercube prior space.
 parraylike
The point in the unit hypercube space
x : The position in the desired hyperparameter space of the point.

km
= None¶

ln_likelihood
(p)[source]¶ Provides a convenient wrapper to the ln likelihood function.
This is implemented in a separate function because of the mild peculiarities of how the pickle module needs to serialise functions, which means that instancemethods (which this would become) can’t be serialised.

loghyperpriors
(p)[source]¶ Calculate the log of the hyperprior distributions at a given point.
 pndarray
The location to be tested.

neg_ln_likelihood
(p)[source]¶ Returns the negative of the loglikelihood; designed for use with minimisation algorithms.
 gpheron Regressor object
The gaussian process to be evaluated.
 parraylike
An array of the hyperparameters at which the model is to be evaluated.
 neg_ln_likelihoodfloat
The negative of the loglikelihood for the Gaussian process

prediction
(new_datum, normalised=False)[source]¶ Produce a prediction at a new point, or set of points.
 new_datumarray
The coordinates of the new point(s) at which the GPR model should be evaluated.
 normalisedbool
A flag to indicate if the input is already normalised (this might be the case if you’re trying to efficiently sample to parameter space). If False the input will be normalised to the same range as the training data.
 prediction meanarray
The mean values of the function drawn from the Gaussian Process.
 prediction variancearray
The variance values for the function drawn from the GP.

rmse
()[source]¶ Calculate the root mean squared error of the whole model.
 rmsefloat
The root mean squared error.

save
(filename)[source]¶ Save the Gaussian Process to a file which can be reloaded later.
 filenamestr
The location at which the Gaussian Process should be written.
In the current implementation the serialisation of the GP is performed by the python pickle library, which isn’t guaranteed to be binarycompatible with all machines.

train
(method='MCMC', metric='loglikelihood', sampler='ensemble', **kwargs)[source]¶ Train the Gaussian process by finding the optimal values for the kernel hyperparameters.
 methodstr {“MCMC”, “MAP”, “nested”}
The method to be employed to calculate the hyperparameters.
 metricstr
The metric which should be used to assess the model.
 hyperpriorslist
The hyperprior distributions for the hyperparameters. Defaults to None, in which case the prior is uniform over all real numbers.

heron.training module¶
These are functions designed to be used for training a Gaussian process made using heron.

heron.training.
cross_validation
(p, gp)[source]¶ Calculate the crossvalidation factor between the training set and the test set.
 gpheron.Regressor
The Gaussian process object.
 parray, optional
The hyperparameters for the Gaussian process kernel. Defaults to None, which causes the current values for the hyperparameters to be used.
 cvfloat
The cross validation of the test data and the model.

heron.training.
ln_likelihood
(p, gp)[source]¶ Returns to loglikelihood of the Gaussian process, which can be used to learn the hyperparameters of the GP.
 gpheron Regressor object
The gaussian process to be evaluated
 parraylike
An array of the hyperparameters at which the model is to be evaluated.
 ln_likelihoodfloat
The loglikelihood for the Gaussian process
TODO Add the ability to specify the priors on each hyperparameter.

heron.training.
run_sampler
(sampler, initial, iterations)[source]¶ Run the MCMC sampler for some number of iterations, but output a progress bar so you can keep track of what’s going on

heron.training.
run_training_map
(gp, metric='loglikelihood', repeats=20, **kwargs)[source]¶ Find the maximum a posteriori training values for the Gaussian Process.
 gpheron.GaussianProcess,
The Gaussian process object.
 metric{“loglikelihood”, “cv”}
The metric to be used to train the MCMC. Defaults to log likelihood (loglikelihood), which is the more traditionally Bayesian manner, but crossvalidation (cv) is also available.
 repeatsint, optional
The number of times that the optimisation should be repeated in order to partially combat having the optimiser choose a local rather than the global maximum log_like.
The current implementation has no way of specifying the optimisation algorithm.
TODO Add an option to change the optimisation algorithm.

heron.training.
run_training_mcmc
(gp, walkers=200, burn=500, samples=1000, metric='loglikelihood', samplertype='ensemble')[source]¶ Train a Gaussian process using an MCMC process to find the maximum evidence.
 gpheron.Regressor
The Gaussian process object.
 walkersint
The number of MCMC walkers.
 burnint
The number of samples to be used to evaluate the burnin for the MCMC.
 samplesint
The number of samples to be used for the production sampling.
 metric{“loglikelihood”, “cv”}
The metric to be used to train the MCMC. Defaults to log likelihood (loglikelihood), which is the more traditionally Bayesian manner, but crossvalidation (cv) is also available.
 samplertypestr {“ensemble”, “pt”}
The sampler to be used on the model.
 probsarray
The log probabilities.
 samplesarray
The array of samples from the sampling chains.
At present the algorithm assigns the median of the samples to the value of the kernel vector; this may not ultimately be the best way to do this, and so it should be possible to specify the desired value to be used from the distribution.
TODO Add ability to change median to other statistics for training

heron.training.
run_training_nested
(gp, method='multi', maxiter=None, npoints=1000)[source]¶ Train the Gaussian Process model using nested sampling.
 gpheron.Regressor
The Gaussian Process object.
 method{“single”, “multi”}
The nested sampling method to be used.
 maxiterint
The maximum number of iterations which should be carried out on the marginal likelihood. Optional.
 npointsint
The number of livepoints to use in the optimisation.