ESPAI Resource Hub

Welcome to the ESPAI Resource Hub. Here you’ll find datasets, plots, source code and documentation produced by the project. Resources include training and validation metrics, inference visualisations, and synthetic datasets generated with ESPAI GenAI model, alongside the core software used to produce them. Each release aims to be FAIR—findable, accessible, interoperable, and reusable—so you can reproduce results, benchmark your own methods, and build on this work responsibly.

At a glance: browse the index on the left, open any dataset section for a short overview, downloadable plots and visual summaries (training, validation, inference), file checksums, a tiny CSV preview (first 5 rows), and FAIR artefacts. Status chips like “Available”, “Coming soon”, or “External” indicate exactly where we are in the release process.

Datasets

ESPAI results are grouped into three categories: Training / Validation Metrics (A), Inference Plots (B), and Synthetic Dataset (C).

Need the full files? See “Licence & Citation” for terms and preferred citation, then follow the repository or the contact instructions where noted.

A. Training / Validation Metrics

A1. Training accuracy for each EPIC camera.

Visual summaries of the images. Click an image to download the full-resolution version.

Training accuracy curve over epochs for MOS1
MOS1 - Training accuracy curve.
Training accuracy curve over epochs for MOS2
MOS2 - Training accuracy curve.
Training accuracy curve over epochs for pn
pn - Training accuracy curve.

A2. Validation accuracy for each EPIC camera.

Visual summaries of the images. Click an image to download the full-resolution version.

Validation accuracy curve over epochs for MOS1
MOS1 - Validation accuracy curve.
Validation accuracy curve over epochs for MOS2
MOS2 - Validation accuracy curve.
Validation accuracy curve over epochs for pn
pn - Validation accuracy curve.

A3. Validation statistics for each EPIC camera.

Visual summaries of the images. Click an image to download the full-resolution version.

Validation statistics per epoch graph
MOS1 - Validation statistics curve.
Validation statistics per epoch graph
MOS2 - Validation statistics curve.
Validation statistics per epoch graph
pn - Validation statistics curve.

B. Inference Plots

B1. Plot of inference signal comparison for each EPIC camera.

Visual summaries of the images. Click an image to download the full-resolution version.

Inference signal comparison for MOS1
Inference signal comparison for MOS1 camera Log-scale energy spectra comparison between Signal GTIs and Predicted Signal BTIs.
Inference signal comparison for MOS2
Inference signal comparison for MOS2 Log-scale energy spectra comparison between Signal GTIs and Predicted Signal BTIs.
Inference signal comparison for PN
Inference signal comparison for PN Log-scale energy spectra comparison between Signal GTIs and Predicted Signal BTIs.

C. Synthetic Dataset

C1. Synthetic EPIC-MOS1 Camera Flare events

Overview. This collection contains synthetic solar-flare events generated with the ESPAI Variational Autoencoder (VAE) GenAI model, designed to emulate EPIC-MOS1 camera observations. Each event is provided as a .csv file containing the Pulse Invariant (PI) and the detector coordinates (DETX, DETY); Each event is provided as a CSV file containing three features: PI, DETX and DETY; releases will ship with transparent generation settings.

Intended use: solar-flare background modeling, training machine learning algorithms and controlled experiments on cadence/noise.

mos1_generated_distribution_VAE.csv 15.5 MB SHA-256

Last updated:

mos1_generated_distribution_comparison_VAE.png 27.1 KB

Last updated:

Preview (CSV)

First 5 rows from a tiny sample file.

Loading preview…
FAIR artefacts (status)

We’re progressively adding the artefacts below; items marked “Coming soon” will appear in the next releases.

C2. Synthetic EPIC-MOS2 Camera Flare events

Overview. This collection contains synthetic solar-flare events generated with the ESPAI Variational Autoencoder (VAE) GenAI model, designed to emulate EPIC-MOS2 camera observations. Each event is provided as a .csv file containing the Pulse Invariant (PI) and the detector coordinates (DETX, DETY); Each event is provided as a CSV file containing three features: PI, DETX and DETY; releases will ship with transparent generation settings.

Intended use: solar-flare background modeling, training machine learning algorithms and controlled experiments on cadence/noise.

mos2_generated_distribution_VAE.csv · 26.4 MB SHA-256

Last updated:

mos2_generated_distribution_comparison_VAE.png · 30.1 KB

Last updated:

Preview (CSV)

First 5 rows from a tiny sample file.

Loading preview…
FAIR artefacts (status)

We’re progressively adding the artefacts below; items marked “Coming soon” will appear in the next releases.

C3. Synthetic EPIC-pn Camera Flare events

Overview. This collection contains synthetic solar-flare events generated with the ESPAI Variational Autoencoder (VAE) GenAI model, designed to emulate EPIC-pn camera observations. Each event is provided as a .parquet file containing the Pulse Invariant (PI) and the detector coordinates (DETX, DETY); Each event is provided as a parquet file containing three features: PI, DETX and DETY; releases will ship with transparent generation settings.

Intended use: solar-flare background modeling, training machine learning algorithms and controlled experiments on cadence/noise.

pn_generated_distribution_VAE.parquet · 36.3 MB SHA-256

Last updated:

pn_generated_distribution_comparison_VAE.png · 28.3 KB

Last updated:

Preview (parquet)

First 5 rows from a tiny sample file.

Loading preview…
FAIR artefacts (status)

We’re progressively adding the artefacts below; items marked “Coming soon” will appear in the next releases.

Source Code

Heads up: FAIR artefacts are being published in stages. Items marked “Coming soon” will appear in the next updates; “External” links point to project-controlled sources (e.g., GitHub) when appropriate.

ESPAI Core Repository

  • Languages: Python (PyTorch), JAX (planned)
  • Licence: MIT (TBC)
  • Visibility: Private (opening soon)

View on GitHub Download ZIP

Planned contents
  • Automated data retrieval and SAS processing pipelines
  • Data loaders, preprocessing, and feature engineering utilities
  • Training and evaluation scripts for solar-flare classification models
  • Generative models (KDE / VAE / Autoencoders) for synthetic data generation
  • Statistical validation tools and reproducible configurations

Diagrams & Technical Notes

  • Graphical Description of the VAE
    • Format: PNG diagram
    • Version: v1

    Schematic illustration of the Variational Autoencoder (VAE) pipeline for synthetic data generation.

Licence & Citation

To support ethical reuse and proper attribution, ESPAI provides default licensing and citation templates for datasets and software. Important: if a dataset or repository includes its own LICENSE, citation.txt, or DOI, that local file overrides the defaults below. Always prefer the per-item files when present.

If you adapt the datasets or code, indicate changes and, where practical, link back to this hub so others can find the original materials.

Datasets — Licence & how to cite

Licence (default): Creative Commons Attribution 4.0 International (CC BY 4.0). You must provide appropriate credit and indicate if changes were made. Read the licence.

Recommended dataset citation (plain text)

ESPAI Project (2025). ESPAI Solar Flare — Synthetic Solar Flare Dataset , v1.0.
Koexai. URL: https://ESPAI.koexai.com/resources/  Licence: CC BY 4.0.

Dataset BibTeX (template)

@dataset{ESPAI_C_v0_1_2025,
      author  = {Koexai Srl},
      title   = {ESPAI Solar Flare — Synthetic Solar Flare Dataset},
      year    = {2025},
      version = {1.0},
      url     = {https://ESPAI.koexai.com/resources/},
      license = {CC BY 4.0},
      note    = {}
    }

Tip: if a dataset provides its own citation.txt or DOI, please use that instead of the template above.

Software — Licence & how to cite

Licence (intended): MIT Licence (to be confirmed in the repository). A copy of the licence will be included as LICENSE in the repo. About MIT.

Recommended software citation (plain text)

ESPAI Project (2025). ESPAI Core (v1.0) — Generative models and classification tools.
Source code. URL: https://ESPAI.koexai.com/resources/  Licence: MIT.

Software BibTeX (template)

@software{ESPAI_core_v0_1_2025,
      author  = {Koexai},
      title   = {ESPAI Core},
      year    = {2025},
      version = {1.0},
      url     = {https://ESPAI.koexai.com/resources/},
      license = {MIT},
      note    = {}
    }