config.json

{
  "_ilex": {
    "architecture": "ilex.models.krakencoder.model.Krakencoder",
    "constructor_kwargs": {
      "input_size_list": [
        256
      ],
      "latent_normalize": true,
      "latentsize": 128,
      "linear_include_bias": true
    },
    "flavor_name_list": [
      "SCsdstream_coco439_sift2volnorm"
    ],
    "format": "ilex",
    "framework_version": {
      "equinox": "0.13.8",
      "ilex": "0.0.0.dev0",
      "jax": "0.10.0",
      "jaxlib": "0.10.0",
      "numpy": "2.4.4",
      "safetensors": "0.7.0"
    },
    "has_state": false,
    "origin": "ilex-native",
    "weights_status": "bundled"
  },
  "authors": "Keith W. Jamison, Zijin Gu, Qinxin Wang, Ceren Tozlu, Mert R. Sabuncu, Amy Kuceyeski",
  "copyright": "Network architecture, training code, and pretrained weights -- copyright (c) 2024 Keith W. Jamison; released under the MIT License. JAX / Equinox port code -- copyright (c) the ilex authors, released under the Apache-2.0 / GPL-3.0 dual license used by ilex itself.",
  "data_type": "numpy",
  "description": "Krakencoder (Jamison, Gu, Wang, Tozlu, Sabuncu, Kuceyeski, *Nature Methods* 2025), ported to JAX / Equinox from the upstream PyTorch release (github.com/kjamison/krakencoder). A linked autoencoder that simultaneously bidirectionally translates between structural and functional brain connectivity across different atlases and processing variants ('flavors') via a common 128-dim L2-normalised latent representation. The Nature Methods 2025 publication's canonical model jointly encodes 15 flavors (3 atlases \u00d7 {3 functional connectivity types + 2 structural tractography types}) and maps each to / from the shared latent.\nArchitecture (per published recipe): per-flavor 256-dim PCA input transformation -> 256 -> 128 Linear encoder -> 128-dim L2-normalised latent -> 128 -> 256 Linear decoder -> inverse PCA to the destination flavor's full-dim connectivity space. v0 of this port ships the canonical bundle plus its 15-flavor PCA stack (separate ``krakencoder_pca_stack`` bundle that the ``KrakencoderPipeline`` co-loads).",
  "equinox_version": "0.13.8",
  "ilex_version": "0.0.0.dev0",
  "image_classes": "Brain connectivity matrices in any of the 15 supported flavors (per the canonical bundle's catalog row). Inputs are upper-triangular vectors -- one per flavor per subject. Atlas edge counts: FS86 = 3,655; Shen268 = 35,778; Coco439 = 96,141 (these are the full-dim sizes the per-flavor PCAs project from / inverse-project to).",
  "intended_use": "Inference-time bidirectional translation between brain connectivity flavors via the shared latent representation. Pipeline:\n1. Caller provides input connectivity matrices (upper-triangular\n   vectors) in one or more of the 15 supported flavors.\n2. ``KrakencoderPipeline`` runs each per-flavor PCA forward\n   (full-dim -> 256-dim), encodes to the shared 128-dim\n   L2-normalised latent, then optionally decodes to one or more\n   target flavors and inverse-PCAs back to full dim.\n\nCommon use cases:\n* Translate input(s) in some flavors to predictions in others\n  (the Nature Methods paper's primary application).\n* Extract the 128-dim shared latent as a subject-level\n  representation for downstream prediction tasks (the\n  ``.encode()`` shortcut).\n* Fusion: average multiple input flavors' latents to improve\n  predictive identifiability.\n\nTrained on 700 subjects from the NIH Human Connectome Project; generalises to held-out HCP subjects and (per the paper) extends to non-HCP cohorts with some accuracy degradation.",
  "jax_version": "0.10.0",
  "network_data_format": {
    "inputs": {},
    "outputs": {}
  },
  "numpy_version": "2.4.4",
  "pred_classes": "Outputs match the input shape: predicted connectivity matrices in the requested target flavor(s), as upper-triangular vectors at each flavor's atlas-determined edge count. The ``.encode()``-only shortcut emits a single (128,) L2-normalised latent per input subject.",
  "references": [
    "Jamison K.W., Gu Z., Wang Q., Tozlu C., Sabuncu M.R., Kuceyeski A. (2025). Krakencoder: a unified brain connectome translation and fusion tool. Nature Methods. DOI: 10.1038/s41592-025-02706-2.",
    "Preprint: bioRxiv 10.1101/2024.04.12.589274.",
    "Upstream code: github.com/kjamison/krakencoder (model.py + fetch.py + per-flavor PCA transforms hosted on OSF: osf.io/dfp92)."
  ],
  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20240725.json",
  "task": "brain-connectome translation (15 modality/atlas flavors via a shared latent space)",
  "version": "0.0.0"
}