Kernels¶

base kernel classes¶

class pyqrse.kernels.basekernels.QRSEKernelBase

Bases: object

Unnormalized computational kernel for the QRSE model.

use_entropy¶

1 if uses entropy and 0 if it does not

Type:	int

use_xi¶

True if uses xi and False if it does not

Type:	bool

name¶

short name (S-QRSE) of the kernel (changable)

Type:	str

long_name¶

longer name of the kernel (Symmetric QRSE). Both name and long_name can be changed for chart making purposes. They have no other effects

Type:	str

xi¶

the mean of the data. By default it is set to 0.

Type:	float

pnames¶

list of the parameter names including appropriate label specific subscripts

Type:	list(str)

pnames_latex¶

list of the parameter names for Latex including appropriate label specific subscripts

Type:	list(str)

actions

list of the QRSE action labels

new actions must be of the form of a list of string labels that is the same length as the existing list of actions.

code: QRSEModel Identification code for the Kernel

denorm_params(params)

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

generic_actions = ['a0', 'a1']

classmethod getcode()

QRSEModel Identification code for the Kernel

Returns:	string code

classmethod getktype()

QRSEModel Kernel Type

Returns:	string kernel type. Either ‘binary’ or ‘ternary’

indif(params)

The point of indifference between actions.

For binary kernels:

x s.t. p(a0|x)=p(a1|x)

For ternary kernels:

x s.t. p(a0|x)=p(a2|x)

Parameters:	params (np.array([floats]) – parameter values of the model
Returns:	indifference point
Return type:	float

kernel(x, params)

value unnormalized kernel function

kernel = exp(potential + entropy)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

ktype = 'binary'

log_kernel(x, params)

Log of the unnormalized kernel function

log_kernel = potential + entropy

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

n_actions: length of the list of actions

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.basekernels.QRSEKernelBaseBinary: Bases: pyqrse.kernels.basekernels.QRSEKernelBase

class pyqrse.kernels.basekernels.QRSEKernelBaseTernary: Bases: pyqrse.kernels.basekernels.QRSEKernelBase

binary kernels¶

This is the binary kernel docstring!

yep

class pyqrse.kernels.binarykernels.SQRSEKernel

Bases: pyqrse.kernels.basekernels.QRSEKernelBaseBinary

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

indif(params)

The point of indifference between actions.

For binary kernels:

x s.t. p(a0|x)=p(a1|x)

For ternary kernels:

x s.t. p(a0|x)=p(a2|x)

Parameters:	params (np.array([floats]) – parameter values of the model
Returns:	indifference point
Return type:	float

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.SQRSEKernelNoH

Bases: pyqrse.kernels.binarykernels.SQRSEKernel

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

class pyqrse.kernels.binarykernels.SFQRSEKernel

Bases: pyqrse.kernels.binarykernels.SQRSEKernel

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.SFCQRSEKernel

Bases: pyqrse.kernels.binarykernels.SFQRSEKernel

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.ABXQRSEKernel

Bases: pyqrse.kernels.binarykernels.SFQRSEKernel

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.ABXQRSEKernelNH

Bases: pyqrse.kernels.binarykernels.SFQRSEKernel

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.ABXCQRSEKernel

Bases: pyqrse.kernels.binarykernels.ABXQRSEKernel

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.binarykernels.ABQRSEKernel

Bases: pyqrse.kernels.binarykernels.ABXQRSEKernel

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

ternary kernels¶

class pyqrse.kernels.ternarykernels.AAQRSEKernel

Bases: pyqrse.kernels.basekernels.QRSEKernelBaseTernary

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

indif(params)

The point of indifference between actions.

For binary kernels:

x s.t. p(a0|x)=p(a1|x)

For ternary kernels:

x s.t. p(a0|x)=p(a2|x)

Parameters:	params (np.array([floats]) – parameter values of the model
Returns:	indifference point
Return type:	float

log_kernel(x, params)

Log of the unnormalized kernel function

log_kernel = potential + entropy

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.ternarykernels.ATQRSEKernel

Bases: pyqrse.kernels.ternarykernels.AAQRSEKernel

entropy(x, params)

Entropy of conditional action distribution

H(p(a|x)) = SUM p(a_i|x) for i=1,2 (binary) (i=1,2,3 for ternary)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

log_kernel(x, params)

Log of the unnormalized kernel function

log_kernel = potential + entropy

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

logits(x, params)

The probability distribution of agent actions at a given value of x.

This also referred to as the conditional action distribution given x.

For instance:

binary_logits = p(a0|x), p(a1|x)

ternary_logits = p(p0|x), p(a1|x), p(a2|x)

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	tuple(float) or tuple(np.array([float]))

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.ternarykernels.AAXQRSEKernel

Bases: pyqrse.kernels.ternarykernels.AAQRSEKernel

potential = -b*(p_buy - p_sell)*(x-xi)

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.ternarykernels.AXQRSEKernel

Bases: pyqrse.kernels.ternarykernels.AAQRSEKernel

log_kernel(x, params)

Log of the unnormalized kernel function

log_kernel = potential + entropy

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

potential(x, params)

potential function of the kernel

Parameters:	x (float or np.array([float]) – value of data being tested params (np.array([float])) – array of parameter values
Returns:	float or np.array([float])

set_params0(data=None, weights=None)

Initial parameter value set based on data

Parameters:	data weights
Returns:	np.array([float])

class pyqrse.kernels.ternarykernels.AQRSEKernel: Bases: pyqrse.kernels.ternarykernels.AXQRSEKernel

Kernel Module contents¶

QRSE Kernels¶

Base Classes - pyqrse.kernels.base¶

QRSEBaseKernel Abtract Base For All QRSE Kernels QRSEKernelBaseBinary Abtract Base For All Binary QRSE Kernels QRSEKernelBaseTernary Abtract Base For All Ternary QRSE Kernels

Binary Action Kernels - pyqrse.kernels.binary¶

SQRSEKernel Symmetric QRSE Kernel SQRSEKernelNoH Symmetric QRSE (NO Entropy Term) SFQRSEKernel Scharfenaker and Foley QRSE SFCQRSEKernel Scharfenaker and Foley QRSE (Centered) ABQRSEKernel Asymmetric-Beta QRSE ABCQRSEKernel Asymmetric-Beta QRSE (Centered)

Ternary Action Kernels - pyqrse.kernels.ternary¶

AAQRSEKernel Asymmetric-Action QRSE AAXQRSEKernel Asymmetric-Action(xi) QRSE ATQRSEKernel Asymmetric-Temperature QRSE