{
  "docId": "019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "docSlug": "bf350dd574c19997",
  "documentTitle": "2021 Air Street Capital The State of AI Report 2021",
  "authorId": "AirStreetCapital",
  "authorName": "Air Street Capital",
  "documentKindSlug": "consulting-deck",
  "documentKindLabel": "Consulting deck",
  "sourceTypeSlug": "vc_research",
  "sourceTypeLabel": "VC research",
  "presentationDate": null,
  "orientation": "landscape",
  "aspectRatio": 1.777,
  "pageNumber": 67,
  "pageCount": 188,
  "prevPage": 66,
  "nextPage": 68,
  "slideType": "industry_trends",
  "function": "diagnose",
  "density": "overcrowded",
  "nDataPoints": 12,
  "notes": "The chart shows a bubble plot comparing ROC-AUC score vs GPU Memory (GB) for various GNN models.",
  "elementsJson": [
    "headline_text",
    "paragraph",
    "bullet_list",
    "bubble_chart"
  ],
  "metadataConfidence": 0.95,
  "imagePath": null,
  "slideHref": "/slides/019dd923-5e88-73ef-bd5c-fd5012384ae3/67",
  "deckHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3",
  "deckJsonHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3.json",
  "deckAnchorHref": "/decks/019dd923-5e88-73ef-bd5c-fd5012384ae3#slide-67",
  "components": [
    {
      "bbox": null,
      "kind": "callout",
      "text": "RevGNNs outperform existing models with significantly less memory consumption.",
      "attrs": null,
      "subkind": null,
      "toolName": "Visual emphasis",
      "toolSlug": "visual-emphasis",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d2b-1183adae4928",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.35,
        "x": 0.62,
        "y": 0.38
      },
      "kind": "chart",
      "text": "Bubble chart showing ROC-AUC vs GPU Memory (GB) for various GNN models.",
      "attrs": null,
      "subkind": "bubble",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "21ffa9d6-b0c1-4684-b1eb-39b9be09ef05",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.45,
        "w": 0.55,
        "x": 0.05,
        "y": 0.38
      },
      "kind": "list",
      "text": "To overcome the memory bottleneck of large GNNs, we either need new hardware or model architectures that consume less memory.\nA method called deep reversible architectures (RevGNN) offers memory consumption that is independent of the number of layers in a model. RevGNN has a very large capacity at low memory cost and only slightly increased training time compared to baseline GNNs (ResGNN). Their deepest model, RevGNN-Wide, is the deepest GNN to date with 1000 layers.\nWith only a fraction of the memory footprint, RevGNNs outperform some baselines on a node prediction benchmark task. But depth still doesn't help in most tasks, which is worthy of future investigation.",
      "attrs": null,
      "subkind": "bullet",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "c6078be2-ba03-4a07-a6a2-60e2ebbf534e",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": null,
      "kind": "metric",
      "text": "GPU Memory (GB)",
      "attrs": null,
      "subkind": "primary",
      "toolName": "Quantification",
      "toolSlug": "quantification",
      "confidence": null,
      "componentId": "019dd952-47c3-7407-8d2b-1594c4607bd8",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.15,
        "w": 0.55,
        "x": 0.05,
        "y": 0.21
      },
      "kind": "paragraph",
      "text": "While very expressive and powerful, GNN model size doesn't scale well alongside dataset size due to the complexity of modelling millions of nodes and billions of connections. This is problematic for real-world problems when deploying large GNNs for equally large graph datasets without sacrificing model parameters.",
      "attrs": null,
      "subkind": "paragraph",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "88beabca-f530-4091-8ab3-93ecf6f360e3",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.35,
        "x": 0.62,
        "y": 0.82
      },
      "kind": "source-note",
      "text": "Figure: RevGNNs outperform existing models with significantly less memory consumption.",
      "attrs": null,
      "subkind": null,
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "93fac709-7166-4dad-af3d-66a6ed89809e",
      "frameworkName": null,
      "frameworkSlug": null
    },
    {
      "bbox": {
        "h": 0.05,
        "w": 0.9,
        "x": 0.05,
        "y": 0.14
      },
      "kind": "title",
      "text": "Graph Neural Networks: improving the memory and parameter efficiency of large models",
      "attrs": null,
      "subkind": "headline",
      "toolName": null,
      "toolSlug": null,
      "confidence": null,
      "componentId": "8a7d637f-4d37-41a7-add6-832f054949c2",
      "frameworkName": null,
      "frameworkSlug": null
    }
  ],
  "metrics": [],
  "tools": [],
  "frameworks": [],
  "arcBeats": [
    {
      "to": 152,
      "from": 5,
      "beatId": "019dd95a-0682-776c-8e35-0affbadec38e",
      "arcName": "The Triple Take",
      "arcSlug": "triple-take",
      "beatName": "The Facts (What)",
      "beatSlug": "triple-take-the-facts-what",
      "evidence": "Definitions p.5-6, Exec Summary p.7, then Sections 1-3 catalog evidence.",
      "position": 1,
      "confidence": 78,
      "parentBeatName": "Setup",
      "parentBeatSlug": "setup"
    },
    {
      "to": 152,
      "from": 10,
      "beatId": "019dd95a-0682-776c-8e35-1b3ee3a1c012",
      "arcName": "The Mountain",
      "arcSlug": "mountain",
      "beatName": "Rising Action",
      "beatSlug": "mountain-rising-action",
      "evidence": "Three sections accumulating evidence of accelerating progress.",
      "position": 2,
      "confidence": 55,
      "parentBeatName": "Development",
      "parentBeatSlug": "development"
    }
  ],
  "loops": [
    {
      "to": 69,
      "from": 64,
      "name": "Pattern Hunter",
      "slug": "02-pattern-hunter",
      "bestFor": "Time-pressed audiences, building consensus, when data is strong",
      "matchId": "019dd95a-07fe-70ce-8d3d-627b31dfbe04",
      "evidence": "GNNs jump to 4th hottest keyword + multiple application slides (mesh sim, ETA, RevGNN, OGB).",
      "position": 9,
      "objective": "Show GNNs becoming mainstream",
      "structure": "Evidence A -> Evidence B -> Evidence C -> Pattern/Conclusion",
      "confidence": 78,
      "description": "Group multiple pieces of evidence that together point to a pattern or conclusion"
    }
  ],
  "imagePathAlt": null,
  "thumbSrc": null,
  "thumbSrcAlt": null,
  "locked": true
}