Introduction; Geometric and topological perspective on data analysis; Data representations; Learning on point clouds and graphs; Joint data analysis.

Lecture Slides: Intro

Reading:

Visual data sets: ImageNet and ShapeNet; Techniques for annotation and annotation transport.

Lecture Slides: DataSets

Reading: ImageNet, ShapeNet, Annotation1, Annotation2, PartNet

Linear algebraic techniques: principal components analysis (PCA), Kernel PCA.

Lecture Slides: PCA

Reading: PCA Tutorial, KPCA

Linear algebraic techniques: canonical correlation analysis (CCA). Multidimensional scaling (MDS).

Lecture Slides: CCA_MDS

Reading: CCA Tutorial, CCA2, MDS1, MDS2

Homework 1 out.

Graph methods; spectral approaches, graph Laplacians, Laplacian embeddings, spectral clustering.

Lecture Slides: Spectral_Graph

Reading: Spectral graph theory Yale course (first few lectures); spectral clustering tutorial

Non-linear dimensionality reduction: locally linear embeddings, Laplacian eignemaps, Isomap, t-SNE.

Lecture Slides:NLDR

Reading: Isomap, LE, LLE, t-SNE

Computational topology: topology review, complexes, homology groups.

Lecture Slides:TDA

Reading: Topology and Data

Persistent homology, barcodes and persistence diagrams.

Reading: Barcodes, Persistent Homology, Ripser

Lecture Slides: Persistence

Homework 1 due. Homework 2 out.

Topological inference; the Mapper algorithm. Applications.

Lecture Slides: TDA_Applications

Reading: Shape barcodes, Mapper, scalar fields, ToMATo, Time Series

Representations of 3D Geometry: Voxel-Grids, Point Clouds, Meshes and Other Boundary Models, Solid Models.

Lecture Slides: 3D Reps

Reading: Old survey

Geometry processing; Laplace-Beltrami and other operators on meshes.

Lecture Slides: Laplace-Beltrami

Reading: LB1, LB2, ShapeDNA

Rigid and non-rigid shape alignment. Global and local shape descriptors; intrinsic descriptors, heat and wave kernel signatures.

Lecture Slides: Shape_Correspondences

Reading: ICP; RANSAC; Shape descriptors for retrieval; global point signatures; heat kernel signatures; ShapeGoogle

Homework 2 due. Homework 3 out.

Class Midterm

Geometric deep learning; Volumetric and multi-view CNNs for 3D geometry

Lecture Slides: MultiView_Volumetric_DeepNets

Reading: MVCNN1, MCCNN2, VoxelCNN1, VoxelCNN2

Memorial day holiday -- no class

Deep nets for pointclouds and applications to classification and segmentation.

Lecture Slides: PointNets

Reading: PointNet, PointNet++, VoteNet, FlowNet3D

Homework 3 due. Homework 4 out.

Functional spaces and functional maps, variations; map visualization.

Lecture Slides: Functional Maps

Reading: functional maps paper; map visualization; Siggraph 17 course notes

Networks of shapes and images; cycle consistency; map processing and latent spaces.

Lecture Slides: Functional Map Nets

Reading: multi-latent space co-segmentation, 3D object co-segmentation

Deep nets for graphs and meshes.

Lecture Slides: GraphMeshCNNs

Reading: geodesic, non-euclidean, survey

Encoding shape differences and shape variability. Class summary.

Lecture Slides: Shape Differences

Reading: shape differences, shape_from_differences, StructureNet

Homework 4 due.