RICA replaces ICA's global generative model with local Riemannian geometry, introducing a disentanglement tensor based on the Hessian of the log-likelihood and Ricci curvature to measure pointwise disentanglement, which recovers sources across manifolds in controlled tests.
Title resolution pending
28 Pith papers cite this work, alongside 13,002 external citations. Polarity classification is still indexing.
28
Pith papers citing it
13k
external citations · OpenAlex
citation-role summary
background 2
citation-polarity summary
fields
cs.LG 17 astro-ph.GA 2 astro-ph.CO 1 cs.AI 1 cs.CL 1 cs.CV 1 cs.IR 1 cs.RO 1 math.DS 1 q-bio.OT 1roles
background 2representative citing papers
A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.
J-LAW introduces a coupled latent factor graph that jointly optimizes metric poses, latent states, and landmark embeddings to produce maps that are both metric and actionable for planning.
Rest-frame optical galaxy SEDs from a 16-parameter SPS model are captured by five disentangled VAE latents (mass, young stars, dust, soft/hard ionization); metallicity and age are not independent drivers.
The paper proposes an HRR-based unsupervised method for disentanglement, proves that the unbinding operation induces approximately independent symbol-value pairs, and derives a per-slot capacity bound.
Introduces the directional linear separability measure (LSM) as an asymmetric diagnostic for one-sided affine separability of neural representations.
MEDAL distills manifold embeddings into autoencoders to enable out-of-sample extension and held-out validation of dimension reduction methods.
Applies gentlest ascent dynamics and stable manifold methods to compute domain of attraction boundaries for stable equilibria in synchronous-generator power system models.
Task-aligned supervised geometric stability predicts linear steerability with high accuracy while unsupervised stability detects representational drift earlier and with lower false alarms than CKA or Procrustes.
A GAN framework is trained on EAGLE simulation merger trees to generate new realistic trees for semi-analytic galaxy models at modest computational cost.
Galaxy size-mass relations exhibit double power-law breaks at different pivot masses for quiescent versus bulge-dominated samples, coinciding with AGN activity scales.
New representational levels arise through a five-stage bootstrap driven by explanatory insufficiency, not merely by data, scale or prediction error.
ConvAE-Relay retrieval via source-trained autoencoder latent matching achieves 38.34+/-0.07% relative L2 error on 10x Re shift using only source database, with U-Net at 34.72% and matching quality identified as dominant factor.
Derives an asymptotic equivalent for the Representation Gap in equivariant diffusion models, showing it depends primarily on the intrinsic dimension of the task.
DAR replaces GAP with an attention-based aggregation module retrained jointly with the classifier head to disentangle core from spurious features and outperforms DFR on multiple datasets.
Soft Learning optimally combines heterogeneous ML specialists via cross-validated non-negative least squares, achieving top performance on 70% of 37 datasets with formal guarantees and 72-435x CPU speedups over deep networks.
Disease trajectory embeddings from longitudinal EHR data serve as structural priors to enhance multi-organ IDP representation learning, improving AUC and MAE for disease prediction across 159 conditions in UK Biobank.
LHSD estimates local intrinsic dimension in high-D spaces by spectral filtering of the log-density Hessian via SLQ to isolate zero-curvature tangent directions.
Bayesian-ARGOS is a hybrid frequentist-Bayesian method that discovers equations from limited noisy observations more efficiently than SINDy or bootstrap-ARGOS while adding uncertainty quantification.
VER is a conceptual five-operation diagnostic process for flagging possible explanatory insufficiency in learned representations when predictive performance remains satisfactory.
A new framework grades levels of inference capability in data-driven systems to assess compliance with the EU AI Act definition of AI, illustrated via credit scoring workflows.
A methodological framework of five successive representational levels formalizes how persistent explanatory insufficiency in biological systems motivates reformulating the scientific question.
Coverage regularization in minimal MLPs yields lower prototype reconstruction error and higher specialization than baseline or repulsive losses on 1D data from N=3 to 100.
In one Parkinsonian walker, a feed-forward net trained on M1 PCA coordinates plus occlusal descriptors approximates observed M2 centroids and preserves the Level-4 hierarchy dOC3 < dONL < dOC2.5.
citing papers explorer
-
Disentanglement Beyond Generative Models with Riemannian ICA
RICA replaces ICA's global generative model with local Riemannian geometry, introducing a disentanglement tensor based on the Hessian of the log-likelihood and Ricci curvature to measure pointwise disentanglement, which recovers sources across manifolds in controlled tests.
-
A framework for analyzing concept representations in neural models
A new framework shows concept subspaces are not unique, estimator choice affects containment and disentanglement, LEACE works well but generalizes poorly, and HuBERT encodes phone info as contained and disentangled from speaker info while speaker info resists compact containment.
-
J-LAW: Joint Localization and Actionable World Modeling via Coupled Latent Factor Graphs
J-LAW introduces a coupled latent factor graph that jointly optimizes metric poses, latent states, and landmark embeddings to produce maps that are both metric and actionable for planning.
-
pop-cosmos: Disentangling galaxy properties from observables using data-driven approaches
Rest-frame optical galaxy SEDs from a 16-parameter SPS model are captured by five disentangled VAE latents (mass, young stars, dust, soft/hard ionization); metallicity and age are not independent drivers.
-
Disentanglement with Holographic Reduced Representations
The paper proposes an HRR-based unsupervised method for disentanglement, proves that the unbinding operation induces approximately independent symbol-value pairs, and derives a per-slot capacity bound.
-
A Geometric Measure of Linear Separability for Neural Representations
Introduces the directional linear separability measure (LSM) as an asymmetric diagnostic for one-sided affine separability of neural representations.
-
MEDAL: Manifold Embedding Distillation via Autoencoder Learning
MEDAL distills manifold embeddings into autoencoders to enable out-of-sample extension and held-out validation of dimension reduction methods.
-
Calculating Domain of Attraction Boundary of Power Systems Based on the Gentlest Ascent Dynamics
Applies gentlest ascent dynamics and stable manifold methods to compute domain of attraction boundaries for stable equilibria in synchronous-generator power system models.
-
The Geometric Canary: Predicting Steerability and Detecting Drift via Representational Stability
Task-aligned supervised geometric stability predicts linear steerability with high accuracy while unsupervised stability detects representational drift earlier and with lower false alarms than CKA or Procrustes.
-
A Halo Merger Tree Generation and Evaluation Framework
A GAN framework is trained on EAGLE simulation merger trees to generate new realistic trees for semi-analytic galaxy models at modest computational cost.
-
pop-cosmos: Galaxy size evolution across structural and star-formation classifications in COSMOS-Web
Galaxy size-mass relations exhibit double power-law breaks at different pivot masses for quiescent versus bulge-dominated samples, coinciding with AGN activity scales.
-
Bootstrap Theory of Representational Emergence: Explanatory Insufficiency as a Driver of Representation Learning and World Models
New representational levels arise through a five-stage bootstrap driven by explanatory insufficiency, not merely by data, scale or prediction error.
-
Striding Across Reynolds Numbers: Representation Geometry in Neural PDE Generalisation
ConvAE-Relay retrieval via source-trained autoencoder latent matching achieves 38.34+/-0.07% relative L2 error on 10x Re shift using only source database, with U-Net at 34.72% and matching quality identified as dominant factor.
-
Representation Gap: Explaining the Unreasonable Effectiveness of Neural Networks from a Geometric Perspective
Derives an asymptotic equivalent for the Representation Gap in equivariant diffusion models, showing it depends primarily on the intrinsic dimension of the task.
-
Deep Attention Reweighting: Post-Hoc Attention-Based Feature Aggregation in CNNs for Disentangling Core and Spurious Features under Spurious Correlations
DAR replaces GAP with an attention-based aggregation module retrained jointly with the classifier head to disentangle core from spurious features and outperforms DFR on multiple datasets.
-
Soft Learning
Soft Learning optimally combines heterogeneous ML specialists via cross-validated non-negative least squares, achieving top performance on 70% of 37 datasets with formal guarantees and 72-435x CPU speedups over deep networks.
-
From Trajectories to Phenotypes: Disease Progression as Structural Priors for Multi-organ Imaging Representation Learning
Disease trajectory embeddings from longitudinal EHR data serve as structural priors to enhance multi-organ IDP representation learning, improving AUC and MAE for disease prediction across 159 conditions in UK Biobank.
-
Local Hessian Spectral Filtering for Robust Intrinsic Dimension Estimation
LHSD estimates local intrinsic dimension in high-D spaces by spectral filtering of the log-density Hessian via SLQ to isolate zero-curvature tangent directions.
-
Fast and principled equation discovery from chaos to climate
Bayesian-ARGOS is a hybrid frequentist-Bayesian method that discovers equations from limited noisy observations more efficiently than SINDy or bootstrap-ARGOS while adding uncertainty quantification.
-
Detecting Explanatory Insufficiency in Learned Representations: A Framework for Representational Vigilance
VER is a conceptual five-operation diagnostic process for flagging possible explanatory insufficiency in learned representations when predictive performance remains satisfactory.
-
When Do Data-Driven Systems Exhibit the Capability to Infer?
A new framework grades levels of inference capability in data-driven systems to assess compliance with the EU AI Act definition of AI, illustrated via credit scoring workflows.
-
From Performance to Representational Adequacy: A Representational Bootstrap Framework for Adaptive Biological Systems
A methodological framework of five successive representational levels formalizes how persistent explanatory insufficiency in biological systems motivates reformulating the scientific question.
-
From Latent Space to Training Data: Explainable Specialization in Minimal MLPs
Coverage regularization in minimal MLPs yields lower prototype reconstruction error and higher specialization than baseline or repulsive losses on 1D data from N=3 to 100.
-
From Observed Viability to Internal Predictive Approximation: A Single-Subject Latent-Space Analysis of Gait Dynamics Under Occlusal Constraint
In one Parkinsonian walker, a feed-forward net trained on M1 PCA coordinates plus occlusal descriptors approximates observed M2 centroids and preserves the Level-4 hierarchy dOC3 < dONL < dOC2.5.
-
From Organization to Viability: A Multi-Level Analysis of Gait Dynamics Under Occlusal Constraint
In a single-case PCA analysis of gait under three occlusal settings, OC3 showed the smallest M1–M2 centroid displacement, ONL intermediate, and OC2.5 the largest—an ordering that is representation-dependent and non-causal.
-
Observable Performance Does Not Fully Reflect Adaptive System Organization: A Multi-Level Analysis of Gait Dynamics Under Occlusal Constraint
In one Parkinson's patient, six occlusal probes produce overlapping gait scores and UMAP embeddings, so observable performance does not uniquely identify adaptive system state under VDO constraint.
-
Interpretable experiential learning based on state history and global feedback
A transition graph model with utility and evidence counts learns behaviors from state history and feedback, showing performance comparable to neural networks on Atari Breakout.
-
CRADIPOR: Crash Dispersion Predictor
A rank reduction autoencoder combined with classification predicts numerical dispersion in automotive crash simulations more effectively than random forests when using wavelet or slope signal inputs.