Geometric Deep Learning — Schedule
Week-by-week schedule of Geometric Deep Learning.
The operational schedule for Geometric Deep Learning. Per-cohort dates fill in at intake; the structure below is stable across cohorts.
The single source of truth is _data/apprentissage-geometrique.yml. Edits there flow through this page automatically.
| Week | Title | Pitch | Detail |
|---|---|---|---|
| 01 | Motivations — Why Geometry in Deep Learning | AlphaFold, AlphaGo, equivariant molecular models — none works without geometry-aware architecture. The unifying thesis of geometric deep learning, due to Bronstein et al. 2021. | week 01 → |
| 02 | Group Theory for Deep Learning | The minimum group theory you need to read modern equivariance papers, built from the ground up. | week 02 → |
| 03 | Graph Neural Networks | Message passing as the core operation. Why GNNs are permutation-equivariant by construction. | week 03 → |
| 04 | GNN Variants — GAT, GIN, GraphSAGE | The expressive-power hierarchy and the design choices that drive it. | week 04 → |
| 05 | Spectral Graph Learning | Convolution on graphs via the graph Laplacian. Why ChebNet matters and why most modern GNNs went back to spatial methods. | week 05 → |
| 06 | Learning on Manifolds | Molecular structures, brain scans, single-cell embeddings — data that lives on curved surfaces where Euclidean distance lies. Manifold-aware learning is the fix. | week 06 → |
| 07 | Equivariant and Invariant Networks | Building neural networks that commute with group actions. The cleanest case of design-by-symmetry in modern ML. | week 07 → |
| 08 | Point Cloud Representations | Three architectures for unordered sets of 3D points: PointNet, PointNet++, and EquivariantConv. | week 08 → |
| 09 | Applications | Three case studies: AlphaFold, drug discovery, autonomous-driving perception. | week 09 → |
| 10 | Connections to TDA | Topological data analysis and geometric deep learning are coming together. What each side brings to the other. | week 10 → |
| 11 | Current Research Directions | Where the field is heading: scale, expressivity, foundations. | week 11 → |
| 12 | Final project presentations | Each participant presents a geometric deep-learning system applied to a dataset of their choice. | week 12 → |
Operational notes
- Default timezone: Africa/Lagos (UTC+1). Per-cohort timing negotiated at intake.
- Lab notebooks and problem-set repos live in the cohort GitHub organization.
- The bilingual lecture notes remain the reference text.