Disentanglement

Module: benchmarks.metrics.disentanglement

Source: benchmarks/metrics/disentanglement.py

Overview

Disentanglement metrics for generative models.

This module provides metrics for evaluating disentanglement in generative models, particularly for VAE models with controllable generation capabilities.

Classes

DisentanglementMetric

class DisentanglementMetric

MutualInformationGapMetric

class MutualInformationGapMetric

SeparationMetric

class SeparationMetric

Functions

init

def __init__()

init

def __init__()

init

def __init__()

compute

def compute()

compute

def compute()

compute

def compute()

create_dci_metric

def create_dci_metric()

create_mig_metric

def create_mig_metric()

create_sap_metric

def create_sap_metric()

validate_inputs

def validate_inputs()

validate_inputs

def validate_inputs()

validate_inputs

def validate_inputs()

Module Statistics

• Classes: 3
• Functions: 12
• Imports: 9