ARXIV:2604.24662 · PHYSICS AI · SUBMITTED 28 APR · 15:20 UTC · FRESHNESS STALE

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data

K. Michael Martini · Eslam Abdelaleem · Paarth Gulati · Ilya Nemenman · arXiv

A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

Ship in 2-4 weeks›Score3.0Evidence unverified

Opportunity summary

Pain A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

Evidence 0 refs | 3 sources | 50% coverage

Blocker Evidence unverified

Open Build Read PDF Signal Canvas Track

PROBLEM

A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space. The challenge is that the state variables are not directly observable and must be inferred from raw…

METHOD

Full abstract

Identifying the dynamical state variables of a system from high-dimensional observations is a central problem across physical sciences. The challenge is that the state variables are not directly observable and must be inferred from raw high-dimensional data without supervision. Here we introduce DySIB (Dynamical Symmetric Information Bottleneck) as a method to learn low-dimensional representations of time-series data by maximizing predictive mutual information between past and future observation windows while penalizing representation complexity. This objective operates entirely in latent space and avoids reconstruction of the observations. We apply DySIB to an experimental video dataset of a physical pendulum, where the underlying state space is known. The method, with hyperparameters of the learning architecture set self-consistently by the data, recovers a two-dimensional representation that matches the dimensionality, topology, and geometry of the pendulum phase space, with the learned coordinates aligning smoothly with the canonical angle and angular velocity. These results demonstrate, on a well-characterized experimental system, that predictive information in latent space can be used to recover interpretable dynamical coordinates directly from high-dimensional data.

RESULT

ScienceToStartup currently rates this 3.0/10 on the public viability pass. These results demonstrate, on a well-characterized experimental system, that predictive information in latent space can be used to recover interpretable dynamical coordinates directly from…

WHY NOW

Physics AI moved forward this cycle; last verified April 2026. Public score 3.0/10. Production flags indicate code availability.

Continue into Read for claims, analysis, references, and neighboring papers.

Opportunity summary

Score3.0

PainA method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

Evidence0 refs | 3 sources | 50% coverage

Blockerno shell-level blocker reported

Analysis summary

A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

VerifiedSource: PDF linkedVerifiedPaperPack: citation fields availablePartialProof: unverified proof status

Competitive landscape

A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

Segment

Physics AI

Adoption evidence

No public code link in the paper record yet

Commercial read

3.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

{ "contract_version": "paper-r2", "paper_id": "dbc8c1c5-c38d-479d-8f20-5825aaf82f62", "arxiv_id": "2604.24662", "canonical_route": "/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "active_tab": "synced from current hash by the drawer client", "selected_artifact": "information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "endpoints": { "paper_pack": "/api/v1/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data/paper-pack", "build_passport": "/api/v1/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data/build-passport", "mcp_resource": "sciencetostartup://surfaces/paper-workspace" } }

{ "surface": "paper", "mode": "paper", "query": "Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data", "normalized_query": "2604.24662", "route": "/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "paper_ref": "information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "topic_slug": null, "benchmark_ref": null, "dataset_ref": null }

{ "@context": "https://schema.org", "@graph": [ { "@type": "WebPage", "@id": "https://sciencetostartup.com/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data#webpage", "url": "https://sciencetostartup.com/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "name": "Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data", "description": "A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.", "isPartOf": { "@id": "https://sciencetostartup.com/#website" } }, { "@type": "ScholarlyArticle", "@id": "https://sciencetostartup.com/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data#scholarlyArticle", "headline": "Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data", "description": "A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.", "url": "https://sciencetostartup.com/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data", "sameAs": "https://arxiv.org/abs/2604.24662", "identifier": { "@type": "PropertyValue", "propertyID": "arXiv", "value": "2604.24662" }, "isAccessibleForFree": true, "isPartOf": { "@id": "https://sciencetostartup.com/#website" }, "datePublished": "2026-04-27T16:24:45.000Z", "author": [ { "@type": "Person", "name": "K. Michael Martini" }, { "@type": "Person", "name": "Eslam Abdelaleem" }, { "@type": "Person", "name": "Paarth Gulati" }, { "@type": "Person", "name": "Ilya Nemenman" } ], "additionalProperty": [ { "@type": "PropertyValue", "propertyID": "viabilityScore", "value": 3 }, { "@type": "PropertyValue", "propertyID": "researchDomain", "value": "Physics AI" }, { "@type": "PropertyValue", "propertyID": "commercialReadiness", "value": "code" } ] }, { "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://sciencetostartup.com" }, { "@type": "ListItem", "position": 2, "name": "Physics AI", "item": "https://sciencetostartup.com/topics" }, { "@type": "ListItem", "position": 3, "name": "Information bottleneck for learning the phase space of dynam", "item": "https://sciencetostartup.com/paper/information-bottleneck-for-learning-the-phase-space-of-dynamics-from-high-dimensional-experimental-data" } ] } ] }

Competitive landscape

A method to learn low-dimensional representations of time-series data by maximizing predictive mutual information in latent space.

Segment

Physics AI

Adoption evidence

No public code link in the paper record yet

Commercial read

3.0/10 public viability

Direct

not classified

Adjacent

not classified

Substitute

not classified

Unknown

not classified

Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data

Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data

Claim map

Constellation map

Competitive landscape

Buzz

PDF

REFERENCES

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline

Claim map

Constellation map

Competitive landscape

Buzz

PDF

REFERENCES

Related Papers

Subscribe to the weekly brief

Build artifacts

Brief

Experiment plan

Validation checklist

Scientific founder

Translational engineer

Domain operator

GTM lead

Regulatory/clinical advisor

Timeline