add site/src/routes/paper/iclr-ml4mat/+page.md

extract common elements between paper and extended abstract (references, title, authors) to site/src/routes/paper/+layout.svelte

Author: janosh

scripts/make_api_docs.py

@@ -5,7 +5,7 @@
 
 from matbench_discovery import ROOT, URLs
 
-# Update auto-generated API docs.
+"""Update auto-generated API docs viewable on the site's /api page."""
 
 out_path = f"{ROOT}/site/src/routes/api"
 
@@ -20,13 +20,13 @@
 )
 
 # Tweak lazydocs's markdown output:
-# - remove bold tags since they break inline code
-# - make badges linking to GitHub source code blue with flat style, add alt text
 for path in glob(f"{out_path}/*.md"):
     text = open(path).read()
 
+    # remove bold tags since they break inline code
     text = text.replace("<b>", "").replace("</b>", "")
 
+    # make badges linking to GitHub source code blue with flat style, add alt text
     text = text.replace(
         'src="https://img.shields.io/badge/-source-cccccc?style=flat-square"',
         'src="https://img.shields.io/badge/source-blue?style=flat" alt="source link"',

site/src/app.d.ts

@@ -8,11 +8,22 @@ declare module '*package.json' {
   export default pkg
 }
 
+// model metadata files
 declare module '*metadata.yml' {
   const data: import('$lib').ModelMetadata
   export default data
 }
 
+// paper metadata
+declare module '*frontmatter.yml' {
+  const frontmatter: import('$lib').Frontmatter
+  export = frontmatter
+}
+
+declare module '*references.yaml' {
+  export const references: import('$lib').Reference[]
+}
+
 declare module '*element-counts.json' {
   const map: Record<string, number>
   export default map

site/src/lib/index.ts

@@ -70,3 +70,16 @@ export type Reference = {
   type: string
   ISSN?: string
 }
+
+export type Frontmatter = {
+  title: string
+  subtitle?: string
+  tags: string[]
+  authors: {
+    name: string
+    affiliation: string
+    orcid: string
+  }[]
+  affiliations: string[]
+  date: string
+}

site/src/routes/paper/+layout.svelte

@@ -0,0 +1,66 @@
+<script lang="ts">
+  import { References } from '$lib'
+  import { affiliations, authors, date, subtitle, title } from './frontmatter.yml'
+  import { references } from './references.yaml'
+</script>
+
+<h1>{title}<br /><small>{subtitle}</small></h1>
+
+<address>
+  <span>
+    {@html authors
+      .map((author) => `${author.name}<sup>${author.affiliation}</sup>`)
+      .join(`, `)}
+  </span>
+  <span>
+    {@html affiliations.map((affil, idx) => `${idx + 1}. ${affil}`).join(`<br/>`)}
+  </span>
+  <span style="font-weight: lighter;">{date}</span>
+</address>
+
+<div>
+  <slot />
+</div>
+
+<h2>References</h2>
+
+<References {references} />
+
+<style>
+  address {
+    text-align: center;
+  }
+  address span {
+    margin: 1em;
+    display: block;
+  }
+  div :global(summary) {
+    font-weight: 300;
+  }
+  div :global(summary)::before {
+    content: 'Abstract';
+    font-weight: bold;
+    font-size: larger;
+  }
+  /* references */
+  div :global(sup.ref) {
+    text-transform: capitalize;
+  }
+  /*auto-number HTML headings*/
+  div :global(h2) {
+    counter-increment: h2;
+    counter-reset: h3, h4;
+  }
+  div :global(h2::before) {
+    content: counter(h2);
+    margin-right: 10pt;
+  }
+  div :global(h3) {
+    counter-increment: h3;
+    counter-reset: h4;
+  }
+  div :global(h3::before) {
+    content: counter(h2) '.' counter(h3);
+    margin-right: 10pt;
+  }
+</style>

site/src/routes/paper/+page.md

@@ -1,71 +1,15 @@
----
-title: Matbench Discovery
-subtitle: Can machine learning identify stable crystals given unrelaxed structures?
-tags:
-  - Python
-  - machine learning
-  - materials science
-  - materials discovery
-  - benchmark
-  - ensembles
-  - uncertainty estimation
-authors:
-  - name: Janosh Riebesell
-    orcid: 0000-0001-5233-3462
-    corresponding: true
-    affiliation: 1, 2
-  - name: Rhys Goodall
-    orcid: 0000-0002-6589-1700
-    affiliation: 1
-  - name: Anubhav Jain
-    orcid: 0000-0001-5893-9967
-    affiliation: 2
-  - name: Kristin Persson
-    orcid: 0000-0003-2495-5509
-    affiliation: 2
-  - name: Alpha Lee
-    orcid: 0000-0002-9616-3108
-    affiliation: 1
-affiliations:
-  - Cavendish Laboratory, University of Cambridge, UK
-  - Lawrence Berkeley National Laboratory, Berkeley, USA
-date: Jan 31, 2023
-# To create a PDF from this markdown file, run:
-# ```sh
-# cd paper
-# pandoc path/to/this/file --output paper.pdf --citeproc
-# ```
-# Requires `brew install pandoc` on macOS. Not identical but similar output as the
-# artifact generated by the JOSS GitHub action
-# https://github.com/marketplace/actions/open-journals-pdf-generator
----
-
 <script>
   import MetricsTable from '$figs/metrics-table.svelte'
-  import { references } from './references.yaml'
-  import { References } from '$lib'
   import CumulativeClfMetrics from '$figs/cumulative-clf-metrics.svelte'
   import RollingMaeModels from '$figs/rolling-mae-vs-hull-dist-models.svelte'
   import { browser } from '$app/environment'
 </script>
 
-# {title}<br><small>{subtitle}</small>
-
-<address>
-  <span rel="author">
-    {@html authors
-      .map((author) => `${author.name}<sup>${author.affiliation}</sup>`)
-      .join(`, `)}
-  </span>
-  <span data-rel="affiliation">
-    {@html affiliations.map((affil, idx) => `${idx + 1}. ${affil}`).join(`<br/>`)}
-  </span>
-  <span style="font-weight: lighter;">{date}</span>
-</address>
+<summary>
 
-## Abstract
+We present a new machine learning (ML) benchmark for materials stability predictions named `Matbench Discovery`. A goal of this benchmark is to highlight the need to focus on metrics that directly measure their utility in prospective discovery campaigns as opposed to analyzing models based on predictive accuracy alone. Our benchmark consists of a task designed to closely simulate the deployment of ML energy models in a high-throughput search for stable inorganic crystals. We explore a wide variety of models covering multiple methodologies ranging from random forests to GNNs, and from one-shot predictors to iterative Bayesian optimizers and interatomic potential-based relaxers. We find M3GNet to achieve the highest F1 score of 0.58 and $R^2$ of 0.59 while MEGNet wins on discovery acceleration factor (DAF) with 2.94. Our results provide valuable insights for maintainers of high throughput materials databases to start using these models as triaging steps to more effectively allocate compute for DFT relaxations.
 
-We present a new machine learning benchmark for materials stability predictions called **Matbench Discovery**. The primary goal of this benchmark is to evaluate the effectiveness of machine learning energy models at accelerating the search for inorganic crystals and to determine the optimal methodology for this task. Specifically, we aim to answer the question of whether density functional theory emulators like M3GNet or one-shot predictors like Wrenformer perform better. To make the results easily accessible, we provide an online leaderboard with interactive plots that allow for custom model comparisons and looking at a variety of performance metrics. New models are easily added to our leaderboard, allowing this benchmark to grow into the future. We also make our results easily reproducible by releasing all training and test scripts as well as WandB logs of all our experiments. We hope our results provide valuable insights that motivate researchers in the field of materials discovery and builders of high throughput databases to start using these models as triaging steps to more effectively allocate compute for DFT relaxations.
+</summary>
 
 ## Introduction
 
@@ -195,41 +139,3 @@ M3GNet's performance appears to be particularly affected by poor performance on
 ## Acknowledgements
 
 JR acknowledges support from the German Academic Scholarship Foundation (Studienstiftung) and gracious hosting as a visiting affiliate in the groups of [KP](https://perssongroup.lbl.gov/people) and [AJ](https://hackingmaterials.lbl.gov).
-
-## References
-
-<References {references} />
-
-<style>
-  #abstract,
-  #abstract + p {
-    font-weight: 300;
-  }
-  address {
-    text-align: center;
-  }
-  address span {
-    margin: 1em;
-    display: block;
-  }
-  /*auto-number HTML headings*/
-  h2 {
-    counter-increment: h2;
-    counter-reset: h3, h4;
-  }
-  h2::before {
-    content: counter(h2);
-    margin-right: 10pt;
-  }
-  h3 {
-    counter-increment: h3;
-    counter-reset: h4;
-  }
-  h3::before {
-    content: counter(h2) '.' counter(h3);
-    margin-right: 10pt;
-  }
-  sup.ref {
-    text-transform: capitalize;
-  }
-</style>