TTK base class for representing differentiable topological losses to be used in dimension reduction. More...

#include <TopologicalLoss.h>

Public Types
enum class	REGUL : std::uint8_t { NO_REGUL , TOPOAE , TOPOAE_DIM1 , CASCADE , ASYMMETRIC_CASCADE , W_DIM1 }

Detailed Description

TTK base class for representing differentiable topological losses to be used in dimension reduction.

Author: Mattéo Clémot matte.nosp@m.o.cl.nosp@m.emot@.nosp@m.univ.nosp@m.-lyon.nosp@m.1.fr

Date: July 2024.

This file defines the TopologicalLoss class. It can compute the following differentiable torch::Tensor quantities that can be used as loss function terms for topological regularization in an autoencoder-based dimension reduction technique : (1) : "Topological Autoencoder" loss (2) : 1-dimensional-extended "Topological Autoencoder" loss (3) : cascade-extended "Topological Autoencoder" loss (4) : asymmetric cascade-extended "Topological Autoencoder" loss (5) : Wasserstein distance between 1-dimensional persistence diagrams

See also: TopologicalDimensionReduction.cpp for a usage example.

Definition at line 32 of file TopologicalLoss.h.

Member Enumeration Documentation

◆ REGUL

enum class ttk::TopologicalLoss::REGUL : std::uint8_t

strong

Enumerator
NO_REGUL	no regularization (returns 0)
TOPOAE	"Topological Autoencoder" loss
TOPOAE_DIM1	1-dimensional-extended "Topological Autoencoder" loss
CASCADE	cascade-extended "Topological Autoencoder" loss
ASYMMETRIC_CASCADE	asymmetric cascade-extended "Topological Autoencoder" loss
W_DIM1	Wasserstein distance between 1-dimensional persistence diagrams

Definition at line 34 of file TopologicalLoss.h.

The documentation for this class was generated from the following file:

/home/julien/Pro/git/github/thomas/ttk-dani/core/base/topologicalDimensionReduction/TopologicalLoss.h

Public Types

Detailed Description

Member Enumeration Documentation

◆ REGUL