Abstract:We present $\texttt{LAMINAR}$, a novel unsupervised machine learning pipeline designed to enhance the representation of structure within data via producing a more-informative distance metric. Analysis methods in the physical sciences often rely on standard metrics to define geometric relationships in data, which may fail to capture the underlying structure of complex data sets. $\texttt{LAMINAR}$ addresses this by using a continuous-normalising-flow and inverse-transform-sampling to define a Riemannian manifold in the data space without the need for the user to specify a metric over the data a-priori. The result is a locally-adaptive-metric that produces structurally-informative density-based distances. We demonstrate the utility of $\texttt{LAMINAR}$ by comparing its output to the Euclidean metric for structured data sets.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: **How to more accurately capture and represent the intrinsic structure of data in complex datasets, especially in the geometric relationship analysis methods commonly used in physical science research, where standard metrics (such as Euclidean metrics) may not be able to fully capture the complex structure of data**. Specifically, the author proposes a new unsupervised machine - learning pipeline - LAMINAR (Locally Adaptive Metrics using Invertible Networks on Anisotropic Regions), aiming to enhance the representation of data structure by generating more informative distance metrics. Traditional methods usually rely on Euclidean metrics to define geometric relationships in data, but this may overlook the complex structure of the dataset. LAMINAR defines a Riemannian manifold by using continuous - normalizing - flow and inverse - transform - sampling, without the need for the user to pre - specify the metric. This enables LAMINAR to generate locally - adaptive - metrics, which can better reflect the density and structural characteristics of the data. ### Main problem summary: 1. **Limitations of standard metrics**: Traditional Euclidean metrics may not be able to capture the intrinsic structure of data when dealing with complex datasets. 2. **Need for more effective metric methods**: In order to better preserve and reveal the structural characteristics of data, a metric method that can be adaptively adjusted is required. 3. **No need to pre - define the metric**: Existing density - based methods usually need to pre - define a meaningful metric, while LAMINAR avoids this limitation and provides a more flexible and powerful solution. By solving these problems, LAMINAR can provide better performance in various downstream tasks, such as cluster analysis, etc., and especially performs well in tasks that rely on the representation of data structure.

Learning Locally Adaptive Metrics that Enhance Structural Representation with $\texttt{LAMINAR}$

Learning Exemplar-Represented Manifolds in Latent Space for Classification.

Deep Manifold Computing and Visualization Using Elastic Locally Isometric Smoothness

Markov-Lipschitz Deep Learning

Deep Manifold Computing and Visualization

Adaptive Nonlinear Dimensionality Reduction with a Local Metric.

Neighborhood-Adaptive Structure Augmented Metric Learning

Local Coordinates Alignment (Lca): A Novel Manifold Learning Approach

Local Deep Feature Learning Framework for 3D Shape.

Manifold learning through locally linear reconstruction based on Euclidean distance

Deep Manifold Transformation for Nonlinear Dimensionality Reduction

Curvature Augmented Manifold Embedding and Learning

DLME: Deep Local-flatness Manifold Embedding

Learning Distances from Data with Normalizing Flows and Score Matching

Deep Localized Metric Learning.

Adaptive neighborhood Metric learning

LOCA: LOcal Conformal Autoencoder for standardized data coordinates

Towards Interpretable Deep Metric Learning with Structural Matching

Scalable unsupervised alignment of general metric and non-metric structures

Sailing in high-dimensional spaces: Low-dimensional embeddings through angle preservation