tutorials/tutorial_catalog.yml at main · pykale/tutorials · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
# Source of truth for the tutorial gallery.
#
# `conf.py`, `scripts/validate_catalog.py` and `scripts/export_landing_page.py` all read this file.
# Adding a tutorial means adding a folder plus one line here -- no Sphinx config edit required.
#
# Decision rule (see README): the folder path fixes the owning domain and the task. Everything else
# -- modality, method, fusion, dataset, difficulty -- lives in each tutorial's metadata.yml and
# surfaces as a landing-page filter.

version: 1

buckets:
  - id: protein
    label: Protein
    color: "#0f766e"
    icon: molecule
    status: active
    examples:
      - Drug-target affinity (DeepDTA)
      - Polypharmacy side effects
    summary: >-
      Molecular therapeutics. Broader than protein-only work: drug-target binding and
      polypharmacy side effects both live here.
    tutorials:
      - drug_target_affinity_prediction
      - drug_target_interaction_prediction
      - polypharmacy_side_effect_prediction

  - id: cancer
    label: Cancer
    color: "#7c3aed"
    icon: layers
    status: active
    examples:
      - Multiomics integration (MOGONET)
    summary: >-
      Primary owner for cancer applications. Tutorials here may also be tagged for biomedical or
      neurodegeneration use through their metadata.
    tutorials:
      - multiomics_cancer_classification

  - id: cardiac
    label: Cardiac
    color: "#dc2626"
    icon: pulse
    status: active
    examples:
      - Cardiac MRI diagnosis
      - Landmark uncertainty
    summary: >-
      Cardiac imaging and the uncertainty tooling that grew out of it. The quantile-binning method
      is reusable across domains, but the cardiac tutorials remain its primary owner.
    tutorials:
      - cardiac_mri_diagnosis
      - landmark_uncertainty_estimation

  - id: materials
    label: Materials
    color: "#ea580c"
    icon: lattice
    status: coming_soon
    examples:
      - Structure/property prediction
      - Crystal descriptors
    summary: >-
      Reserved. No PyKale example maps here yet -- earmarked for structure/property prediction,
      crystal and material descriptors, and materials benchmark tasks.
    tutorials: []

  # Domain-agnostic method demos. Labelled "Others" to match the folder (others/) and listed last.
  - id: others
    label: Others
    color: "#2563eb"
    icon: nodes
    status: active
    examples:
      - Domain adaptation on digits
      - Few-shot image classification
    summary: >-
      Domain-agnostic method demonstrations: image and video recognition, domain adaptation,
      few-shot learning and data loading. Start here if you are new to PyKale.
    groups:
      - id: image_classification
        label: Image classification
        tutorials:
          - cifar_cnn_transformer
          - cifar_isonet
          - digit_domain_adaptation
          - office_multisource_adaptation
    tutorials:
      - audio_visual_digit_classification
      - action_recognition
      - fewshot_image_classification
      - multisite_neuroimaging_classification
      - synthetic_domain_adaptation
      - video_data_loading

# Related projects, shown in the homepage ecosystem section. Entries without a `url` render as plain
# text (never a dead link); UKOMAIN and OMAIB have no public tutorial URL yet, so they carry none.
ecosystem:
  - name: PyKale core
    url: https://github.com/pykale/pykale
    description: The library these tutorials are built on -- multimodal and transfer learning.
  - name: PyKale data
    url: https://github.com/pykale/data
    description: Public datasets the tutorials download from.
  - name: PyKale docs
    url: https://pykale.readthedocs.io
    description: API reference for every kale module a tutorial uses.
  - name: UKOMAIN
    description: UK Open Multimodal AI Network -- the community initiative behind this work.
    planned: true
  - name: OMAIB
    description: Open Multimodal AI for Biomedicine -- biomedical applications of PyKale.
    planned: true
  - name: Domain packages
    description: Planned kale-protein, kale-cancer, kale-cardiac and kale-materials extensions.
    planned: true

# Curated routes through the tutorials. Slugs are validated against the catalog by validate_catalog.
learning_paths:
  - id: new_to_pykale
    title: New to PyKale
    summary: Start here. The gentlest path from a first run to a working domain-adaptation model.
    tutorials:
      - synthetic_domain_adaptation
      - digit_domain_adaptation
      - video_data_loading
  - id: transfer_and_domain_adaptation
    title: Transfer & domain adaptation
    summary: PyKale's core strength -- aligning models across shifting data distributions.
    tutorials:
      - digit_domain_adaptation
      - office_multisource_adaptation
      - action_recognition
      - multisite_neuroimaging_classification
  - id: multimodal_learning
    title: Multimodal learning
    summary: Combining modalities -- audio and image, graphs and sequences, multiomics.
    tutorials:
      - audio_visual_digit_classification
      - drug_target_interaction_prediction
      - multiomics_cancer_classification
  - id: biomedical_ai
    title: Biomedical AI
    summary: Applications across protein, cancer and cardiac data.
    tutorials:
      - drug_target_affinity_prediction
      - multiomics_cancer_classification
      - cardiac_mri_diagnosis
      - landmark_uncertainty_estimation