c3.libraries.algorithms¶
Collection of (optimization) algorithms. All entries share a common signature with optional arguments.
Module Contents¶
- c3.libraries.algorithms.single_eval(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
Return the function value at given point.
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) – Algorithm specific options
- c3.libraries.algorithms.grid2D(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
Two dimensional scan of the function values around the initial point.
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) –
Options include points : int
The number of samples
- boundslist
Range of the scan for both dimensions
- c3.libraries.algorithms.sweep(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
One dimensional scan of the function values around the initial point.
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) –
Options include points : int
The number of samples
- boundslist
Range of the scan
- c3.libraries.algorithms.adaptive_scan(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
One dimensional scan of the function values around the initial point, using adaptive sampling
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) –
Options include
- accuracy_goal: float
Targeted accuracy for the sampling algorithm
- probe_listlist
Points to definitely include in the sampling
- init_pointboolean
Include the initial point in the sampling
- c3.libraries.algorithms.tf_sgd(x_init: numpy.ndarray, fun: Callable = None, fun_grad: Callable = None, grad_lookup: Callable = None, options: dict = {}) scipy.optimize.OptimizeResult[source]¶
Optimize using TensorFlow Stochastic Gradient Descent with Momentum https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/SGD
- Parameters
x_init (np.ndarray) – starting value of parameter(s)
fun (Callable, optional) – function to minimize, by default None
fun_grad (Callable, optional) – gradient of function to minimize, by default None
grad_lookup (Callable, optional) – lookup stored gradients, by default None
options (dict, optional) – optional parameters for optimizer, by default {}
- Returns
SciPy OptimizeResult type object with final parameters
- Return type
OptimizeResult
- c3.libraries.algorithms.tf_adam(x_init: numpy.ndarray, fun: Callable = None, fun_grad: Callable = None, grad_lookup: Callable = None, options: dict = {}) scipy.optimize.OptimizeResult[source]¶
Optimize using TensorFlow ADAM https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Adam
- Parameters
x_init (np.ndarray) – starting value of parameter(s)
fun (Callable, optional) – function to minimize, by default None
fun_grad (Callable, optional) – gradient of function to minimize, by default None
grad_lookup (Callable, optional) – lookup stored gradients, by default None
options (dict, optional) – optional parameters for optimizer, by default {}
- Returns
SciPy OptimizeResult type object with final parameters
- Return type
OptimizeResult
- c3.libraries.algorithms.tf_rmsprop(x_init: numpy.ndarray, fun: Callable = None, fun_grad: Callable = None, grad_lookup: Callable = None, options: dict = {}) scipy.optimize.OptimizeResult[source]¶
Optimize using TensorFlow RMSProp https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/RMSprop
- Parameters
x_init (np.ndarray) – starting value of parameter(s)
fun (Callable, optional) – function to minimize, by default None
fun_grad (Callable, optional) – gradient of function to minimize, by default None
grad_lookup (Callable, optional) – lookup stored gradients, by default None
options (dict, optional) – optional parameters for optimizer, by default {}
- Returns
SciPy OptimizeResult type object with final parameters
- Return type
OptimizeResult
- c3.libraries.algorithms.tf_adadelta(x_init: numpy.ndarray, fun: Callable = None, fun_grad: Callable = None, grad_lookup: Callable = None, options: dict = {}) scipy.optimize.OptimizeResult[source]¶
Optimize using TensorFlow Adadelta https://www.tensorflow.org/api_docs/python/tf/keras/optimizers/Adadelta
- Parameters
x_init (np.ndarray) – starting value of parameter(s)
fun (Callable, optional) – function to minimize, by default None
fun_grad (Callable, optional) – gradient of function to minimize, by default None
grad_lookup (Callable, optional) – lookup stored gradients, by default None
options (dict, optional) – optional parameters for optimizer, by default {}
- Returns
SciPy OptimizeResult type object with final parameters
- Return type
OptimizeResult
- c3.libraries.algorithms.lbfgs(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
Wrapper for the scipy.optimize.minimize implementation of LBFG-S. See also:
https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) – Options of scipy.optimize.minimize
- Returns
Scipy result object.
- Return type
Result
- c3.libraries.algorithms.lbfgs_grad_free(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
Wrapper for the scipy.optimize.minimize implementation of LBFG-S. We let the algorithm determine the gradient by its own.
See also:
https://docs.scipy.org/doc/scipy/reference/optimize.minimize-lbfgsb.html
- Parameters
x_init (float) – Initial point
fun (callable) – Goal function
fun_grad (callable) – Function that computes the gradient of the goal function
grad_lookup (callable) – Lookup a previously computed gradient
options (dict) – Options of scipy.optimize.minimize
- Returns
Scipy result object.
- Return type
Result
- c3.libraries.algorithms.cmaes(x_init, fun=None, fun_grad=None, grad_lookup=None, options={})[source]¶
Wrapper for the pycma implementation of CMA-Es. See also:
http://cma.gforge.inria.fr/apidocs-pycma/
- Parameters
x_init (float) – Initial point.
fun (callable) – Goal function.
fun_grad (callable) – Function that computes the gradient of the goal function.
grad_lookup (callable) – Lookup a previously computed gradient.
options (dict) –
Options of pycma and the following custom options.
- noisefloat
Artificial noise added to a function evaluation.
- init_pointboolean
Force the use of the initial point in the first generation.
- spreadfloat
Adjust the parameter spread of the first generation cloud.
- stop_at_convergenceint
Custom stopping condition. Stop if the cloud shrunk for this number of generations.
- stop_at_sigmafloat
Custom stopping condition. Stop if the cloud shrunk to this standard deviation.
- Returns
Parameters of the best point.
- Return type
np.ndarray