models/basic.py

---

pycmtensor.models.basic

BaseModel(**kwargs)

Bases: object

Basic model class object

Attributes:

Name	Type	Description
name	str	name of the model
config	Config	pycmtensor config object
rng	Generator	random number generator
params	list	list of model parameters (betas & weights)
betas	list	list of model scalar betas
sigmas	list	list of model scalar sigmas
weights	list	list of model weight matrices
biases	list	list of model vector biases
updates	list	list of (param, update) tuples
learning_rate	TensorVariable	symbolic reference to the learning rate
results	Results	stores the results of the model estimation

get_weights()

returns the values of the weights

Returns:

Type	Description
dict	weight values

get_biases()

returns the values of the biases

Returns:

Type	Description
dict	biases values

get_betas()

returns the values of the betas

Returns:

Type	Description
dict	beta values

reset_values()

resets the values of all parameters

include_params_for_convergence(*args, **kwargs)

dummy method for additional parameter objects for calculating convergence

Parameters:

Name	Type	Description	Default
*args	None	overloaded arguments	()
**kwargs	dict	overloaded keyword arguments	{}

Returns:

Type	Description
OrderedDict	a dictonary of addition parameters to include

include_regularization_terms(*regularizers)

dummy method for additional regularizers into the cost function

Parameters:

Name	Type	Description	Default
*args	None	overloaded arguments	required

Returns:

Type	Description
list[TensorVariable]	a list of symbolic variables that specify additional regualrizers to minimize against

build_cost_updates_fn(updates)

Builds a function that calculates the cost updates for a model.

Parameters:

Name	Type	Description	Default
updates	dict	A dictionary of updates for the model.	required

Returns:

Type	Description
	None

predict(dataset)

Make predictions on a given dataset.

Parameters:

Name	Type	Description	Default
dataset		The dataset on which predictions are to be made.	required

Returns:

Type	Description
dict	A dictionary containing the predicted choices, true choices, and choice probabilities
dict	for each data point in the dataset.

elasticities(dataset, wrt_choice)

Calculate the elasticities of the model based on the given dataset and choice.

Parameters:

Name	Type	Description	Default
dataset		The dataset used to calculate the elasticities.	required
wrt_choice		The choice with respect to which the elasticities are calculated.	required

Returns:

Type	Description
	The elasticities of the model based on the given dataset and choice.

extract_params(cost, variables) staticmethod

Extracts Param objects from variables

Parameters:

Name	Type	Description	Default
cost	TensorVariable	function to evaluate	required
variables	Union[dict, list]	list of variables from the current program	required

drop_unused_variables(cost, params, variables) staticmethod

Internal method to remove ununsed tensors

Parameters:

Name	Type	Description	Default
cost	TensorVariable	function to evaluate	required
params	Param	param objects	required
variables	dict	list of array variables from the dataset	required

Returns:

Type	Description
list	a list of param names which are not used in the model

compute(model, ds, update=False, **params)

Function for manual computation of model by specifying parameters as arguments

Parameters:

Name	Type	Description	Default
model	BaseModel	model to train	required
ds	Dataset	dataset to use for training	required
update	bool	if True, run cost_updates_fn once before computing the model	False
**params	dict	keyword arguments for model coefficients (Params)	{}

Returns:

Name	Type	Description
dict		model likelihood and error for training and validation datasets

Example:

compute(model, ds, b_time=-5.009, b_purpose=0.307, asc_pt=-1.398, asc_drive=4.178,
        asc_cycle=-3.996, b_car_own=1.4034)

train(model, ds, **kwargs)

Main training loop

Parameters:

Name	Type	Description	Default
model	BaseModel	model to train	required
ds	Dataset	dataset to use for training	required
**kwargs	dict	overloaded keyword arguments. See configuration in the user guide for details on possible options	{}