Built site for gh-pages

Author: Quarto GHA Workflow Runner

.nojekyll

@@ -1 +1 @@
-ccce5cd2
+a0e96805

index-meca.zip

@@ -6,7 +6,7 @@
     <meta name="viewport" content="width=device-width, initial-scale=1.0, user-scalable=yes">
 
     <meta name="author" content="Philipp Pelz">
-    <meta name="dcterms.date" content="2026-07-05">
+    <meta name="dcterms.date" content="2026-07-11">
     <meta name="keywords" content="Machine Learning, Materials Science, Materials Processing, Materials Characterization, Deep Learning, Microstructure Analysis, Process Optimization">
 
     <title>Machine Learning in Materials Processing &amp; Characterization</title>
@@ -166,10 +166,10 @@
 ">
 <meta name="citation_keywords" content="Machine Learning,Materials Science,Materials Processing,Materials Characterization,Deep Learning,Microstructure Analysis,Process Optimization">
 <meta name="citation_author" content="Philipp Pelz">
-<meta name="citation_publication_date" content="2026-07-05">
-<meta name="citation_cover_date" content="2026-07-05">
+<meta name="citation_publication_date" content="2026-07-11">
+<meta name="citation_cover_date" content="2026-07-11">
 <meta name="citation_year" content="2026">
-<meta name="citation_online_date" content="2026-07-05">
+<meta name="citation_online_date" content="2026-07-11">
 <meta name="citation_language" content="en">
 <meta name="citation_reference" content="citation_title=Strategies for the development of volcanic hazard maps in monogenetic volcanic fields: The example of La Palma (Canary Islands);,citation_author=José Marrero;,citation_author=Alicia García;,citation_author=Manuel Berrocoso;,citation_author=Ángeles Llinares;,citation_author=Antonio Rodríguez-Losada;,citation_author=R. Ortiz;,citation_publication_date=2019-07;,citation_cover_date=2019-07;,citation_year=2019;,citation_doi=10.1186/s13617-019-0085-5;,citation_volume=8;,citation_journal_title=Journal of Applied Volcanology;">
 <meta name="citation_reference" content="citation_title=Materials data science: Introduction to data mining, machine learning, and data-driven predictions for materials science and engineering;,citation_author=Stefan Sandfeld;,citation_publication_date=2024;,citation_cover_date=2024;,citation_year=2024;">
@@ -214,7 +214,7 @@ <h1 class="title">Machine Learning in Materials Processing &amp; Characterizatio
                 <div>
             <div class="quarto-title-meta-heading">Published</div>
             <div class="quarto-title-meta-contents">
-              <p class="date">July 5, 2026</p>
+              <p class="date">July 11, 2026</p>
             </div>
           </div>
 
@@ -463,25 +463,24 @@ <h4 data-number="1.3.2.1" class="anchored" data-anchor-id="week-4-from-classical
 Compare classical features vs simple NN-based features for microstructure tasks.</p>
 <hr>
 </section>
-<section id="week-5-neural-networks-for-microstructure-images" class="level4" data-number="1.3.2.2">
-<h4 data-number="1.3.2.2" class="anchored" data-anchor-id="week-5-neural-networks-for-microstructure-images"><span class="header-section-number">1.3.2.2</span> Week 5 – Neural networks for microstructure images</h4>
+<section id="week-5-unsupervised-learning-clustering-autoencoders" class="level4" data-number="1.3.2.2">
+<h4 data-number="1.3.2.2" class="anchored" data-anchor-id="week-5-unsupervised-learning-clustering-autoencoders"><span class="header-section-number">1.3.2.2</span> Week 5 – Unsupervised learning: clustering &amp; autoencoders</h4>
 <p><em>Lecture: Tuesday, 12.05.2026, 14:15-15:45 | Exercise: Thursday, 14.05.2026, 16:15-17:45 (cancelled - Himmelfahrt)</em></p>
 <p><strong>Slides:</strong> <a href="https://pelzlab.science/public_presentations/ml_for_characterization_and_processing/unit05_unsupervised_learning/01_intro.html">Open</a></p>
 <ul>
-<li>CNN intuition: filters as structure detectors.</li>
-<li>Example tasks: phase segmentation, defect detection, porosity identification.</li>
-<li>Overfitting risks with small datasets.</li>
+<li>Clustering materials data without labels: k-means and Gaussian mixtures.</li>
+<li>CNN embeddings as unsupervised feature spaces for phase discovery.</li>
+<li>Autoencoders for compression, denoising, and anomaly detection.</li>
 </ul>
 <p><strong>Summary:</strong></p>
 <ul>
-<li><strong>Convolutional Neural Networks (CNNs)</strong> for materials characterization</li>
-<li>Hierarchical structure detectors: edges → textures → phase morphologies</li>
-<li>Filters and pooling; parameter efficiency vs.&nbsp;MLPs</li>
-<li>Case studies: phase segmentation, defect detection</li>
-<li>Practical challenges: high-resolution, noisy micrographs</li>
+<li><strong>K-means and GMM</strong> for discovering structure in descriptors and spectra; choosing and defending K</li>
+<li><strong>Frozen CNN embeddings</strong> as feature extractors — no labels needed</li>
+<li><strong>Hyperspectral clustering</strong>: flatten datacubes (EELS/EDS), cluster, re-image as phase maps</li>
+<li><strong>Autoencoders</strong>: low-dimensional bottleneck representations for materials images</li>
+<li><strong>Anomaly detection</strong> via reconstruction-error thresholding on nominal data</li>
 </ul>
-<p><strong>Exercise:</strong><br>
-Train a small CNN on microstructure images; analyze failure cases.</p>
+<p><strong>Exercise:</strong> Cluster materials data with k-means and train a convolutional autoencoder; evaluate clustering quality and anomaly detection.</p>
 <hr>
 </section>
 <section id="week-6-data-scarcity-transfer-learning" class="level4" data-number="1.3.2.3">
@@ -646,14 +645,16 @@ <h4 data-number="1.3.5.1" class="anchored" data-anchor-id="week-13-uncertainty-a
 <p><strong>Exercise:</strong> Compare GP regression and NN ensembles for a process-parameter problem.</p>
 <hr>
 </section>
-<section id="week-14-integration-limits-and-reflection" class="level4" data-number="1.3.5.2">
-<h4 data-number="1.3.5.2" class="anchored" data-anchor-id="week-14-integration-limits-and-reflection"><span class="header-section-number">1.3.5.2</span> Week 14 – Integration, limits, and reflection</h4>
+<section id="week-14-course-recap-from-materials-data-to-trustworthy-models" class="level4" data-number="1.3.5.2">
+<h4 data-number="1.3.5.2" class="anchored" data-anchor-id="week-14-course-recap-from-materials-data-to-trustworthy-models"><span class="header-section-number">1.3.5.2</span> Week 14 – Course recap: from materials data to trustworthy models</h4>
 <p><em>Lecture: Tuesday, 14.07.2026, 14:15-15:45 | Exercise: Thursday, 16.07.2026, 16:15-17:45</em></p>
 <p><strong>Slides:</strong> <a href="https://pelzlab.science/public_presentations/ml_for_characterization_and_processing/unit13_reflection/14_reflection.html">Open</a></p>
 <ul>
-<li>Explainability for experimental ML (CAMs, SHAP).</li>
-<li>Why ML fails in real labs.</li>
-<li>Where ML genuinely changes materials processing.</li>
+<li>The course in one map: raw materials data → representations → models → uncertainty and trust.</li>
+<li>Unit-by-unit recap (Units 1–12), each with four checkpoints: what question, what method &amp; equation, when to reach for it and where it breaks, what you must be able to do in the exam.</li>
+<li>Decision guide: twelve real lab scenarios mapped to the right method and unit.</li>
+<li>Two questions before any deployment: “Where did this data come from?” (Units 1–3) and “How wrong can I afford to be?” (Unit 12).</li>
+<li>Exam scope and mini-project rubric.</li>
 </ul>
 <p><strong>Exercise:</strong> Mini-project presentations and critical discussion.</p>
 <hr>

index-preview.html

@@ -9,11 +9,11 @@
     "Course Curriculum and Materials\n",
     "\n",
     "Philipp Pelz (Materials Science and Engineering)  \n",
-    "July 5, 2026\n",
+    "July 11, 2026\n",
     "\n",
     "This course teaches how machine learning can be applied to experimental data from materials processing and characterization. The focus lies on images, spectra, time-series, and processing parameters, and on understanding how physical data formation interacts with learning algorithms. Students learn to build robust, uncertainty-aware ML pipelines for real experimental workflows, avoiding common pitfalls such as data leakage, overfitting, and spurious correlations."
    ],
-   "id": "5410d27f-40c0-4eb4-a61b-18817791a842"
+   "id": "c6b25e1f-ddd6-4678-9f6b-e0fd77a839bb"
   },
   {
    "cell_type": "raw",
@@ -78,7 +78,7 @@
     "}\n",
     "</style>"
    ],
-   "id": "f35d7870-cbdf-41b3-9bcf-882dbbd27506"
+   "id": "29a41b85-4e32-4459-a825-e1a92bf2b5a9"
   },
   {
    "cell_type": "raw",
@@ -108,7 +108,7 @@
     "  <strong>How to use this course site.</strong> Use this page as the central hub for syllabus, lecture structure, reading, notebooks, and course materials. Formal announcements and enrollment remain on StudOn; code and openly shared resources live in the linked GitHub repository.\n",
     "</div>"
    ],
-   "id": "2f3c1b9b-0c9e-4d20-8ded-8f492d0d4820"
+   "id": "aa4f0297-8dff-4d59-9943-c6d43b6729a4"
   },
   {
    "cell_type": "markdown",
@@ -237,26 +237,25 @@
     "\n",
     "------------------------------------------------------------------------\n",
     "\n",
-    "#### Week 5 – Neural networks for microstructure images\n",
+    "#### Week 5 – Unsupervised learning: clustering & autoencoders\n",
     "\n",
     "*Lecture: Tuesday, 12.05.2026, 14:15-15:45 \\| Exercise: Thursday, 14.05.2026, 16:15-17:45 (cancelled - Himmelfahrt)*\n",
     "\n",
     "**Slides:** [Open](https://pelzlab.science/public_presentations/ml_for_characterization_and_processing/unit05_unsupervised_learning/01_intro.html)\n",
     "\n",
-    "- CNN intuition: filters as structure detectors.\n",
-    "- Example tasks: phase segmentation, defect detection, porosity identification.\n",
-    "- Overfitting risks with small datasets.\n",
+    "- Clustering materials data without labels: k-means and Gaussian mixtures.\n",
+    "- CNN embeddings as unsupervised feature spaces for phase discovery.\n",
+    "- Autoencoders for compression, denoising, and anomaly detection.\n",
     "\n",
     "**Summary:**\n",
     "\n",
-    "- **Convolutional Neural Networks (CNNs)** for materials characterization\n",
-    "- Hierarchical structure detectors: edges → textures → phase morphologies\n",
-    "- Filters and pooling; parameter efficiency vs. MLPs\n",
-    "- Case studies: phase segmentation, defect detection\n",
-    "- Practical challenges: high-resolution, noisy micrographs\n",
+    "- **K-means and GMM** for discovering structure in descriptors and spectra; choosing and defending K\n",
+    "- **Frozen CNN embeddings** as feature extractors — no labels needed\n",
+    "- **Hyperspectral clustering**: flatten datacubes (EELS/EDS), cluster, re-image as phase maps\n",
+    "- **Autoencoders**: low-dimensional bottleneck representations for materials images\n",
+    "- **Anomaly detection** via reconstruction-error thresholding on nominal data\n",
     "\n",
-    "**Exercise:**  \n",
-    "Train a small CNN on microstructure images; analyze failure cases.\n",
+    "**Exercise:** Cluster materials data with k-means and train a convolutional autoencoder; evaluate clustering quality and anomaly detection.\n",
     "\n",
     "------------------------------------------------------------------------\n",
     "\n",
@@ -417,15 +416,17 @@
     "\n",
     "------------------------------------------------------------------------\n",
     "\n",
-    "#### Week 14 – Integration, limits, and reflection\n",
+    "#### Week 14 – Course recap: from materials data to trustworthy models\n",
     "\n",
     "*Lecture: Tuesday, 14.07.2026, 14:15-15:45 \\| Exercise: Thursday, 16.07.2026, 16:15-17:45*\n",
     "\n",
     "**Slides:** [Open](https://pelzlab.science/public_presentations/ml_for_characterization_and_processing/unit13_reflection/14_reflection.html)\n",
     "\n",
-    "- Explainability for experimental ML (CAMs, SHAP).\n",
-    "- Why ML fails in real labs.\n",
-    "- Where ML genuinely changes materials processing.\n",
+    "- The course in one map: raw materials data → representations → models → uncertainty and trust.\n",
+    "- Unit-by-unit recap (Units 1–12), each with four checkpoints: what question, what method & equation, when to reach for it and where it breaks, what you must be able to do in the exam.\n",
+    "- Decision guide: twelve real lab scenarios mapped to the right method and unit.\n",
+    "- Two questions before any deployment: “Where did this data come from?” (Units 1–3) and “How wrong can I afford to be?” (Unit 12).\n",
+    "- Exam scope and mini-project rubric.\n",
     "\n",
     "**Exercise:** Mini-project presentations and critical discussion.\n",
     "\n",
@@ -458,7 +459,7 @@
     "- ML-assisted autofocus or EBSD pattern classification.\n",
     "- Multi-modal fusion of images, spectra, and process parameters."
    ],
-   "id": "a2b304cb-c139-4f60-a8f7-b320c6013df7"
+   "id": "ad610cc8-7fea-4836-ba12-46614b000e3a"
   }
  ],
  "nbformat": 4,