Machine Learning for 3D Data
CSE291-I00 - Winter 2018
Schedule and assignments
|
Date
|
Title
|
Topic
|
|
1/9
|
Introduction | overview of the course, logistics, introduction to deep learning |
|
1/11
|
Deep Learning (Basics) | continue the introduction to deep learning, hands-on tutorial on deep learning frameworks. Readings: AlexNet, VGGNet, ResNet, Batch Normalization, Matrix calculus |
|
1/13
|
Assignment 1 | due: Jan 23th, 2018 |
|
1/16
|
Deep Learning (Advanced I) | visualization, matrix calculus, optimization, generalizability solution and explanation to the quiz in class |
|
1/18
|
Deep Learning (Advanced II) | a glimpse of recent progress in deep learning theory Readings: Hessian of Networks, Langevin Dynamics of SGD, Norm, Sharpness, PAC-Bayesian, Regularization, Generalization |
|
1/23
|
Geometry Foundations | various 3d representations, basics of differential geometry Readings: Polygon Mesh Processing |
|
1/25
|
Geometry Foundations II | curve, surface, geometry processing |
|
1/30
|
Deep Learning for Point Clouds | Earth Mover's Distance, point cloud generation network |
|
2/1
|
Deep Learning for Point Clouds, Embedding | point cloud generation network, embedding theorems, MDS, homotopy |
|
2/4
|
Assignment 2 | due: Feb 13th, 2018 |
|
2/6
|
Deep Learning for Point Clouds, 3D Detection | pointnet series |
|
2/8
|
3D Parameteric Models | 3d parameteric models, shape editing, as-rigid-as-possible, body and face modeling |
|
2/13
|
Recursive Neural Networks | Guest lecture by Prof. Garrison Cottrell |
|
2/15
|
Deep Learning on Multi-view Representation | multi-view CNN, multi-view segmentation, multi-view reconstruction, novel-view synthesis |
|
2/20
|
Deep Learning on Volumetric Representation | 3d cnns, view consistency |
|
2/21
|
Assignment 3 | due: Mar 7th, 2018 |
|
2/22
|
3D GANs | 3D GANs |
|
2/25
|
Project Proposal | due: Mar 1st, 2018 |
|
2/27
|
Graph Laplacian | Basic theory and algorithms of graph laplacian, algebraic connectivity |
|
3/1
|
Graph Laplacian | continue the last lecture |
|
3/6
|
Intrinsics, Isometric Matching | Intrinsics, shape correspondence, Laplace-Bertrami operator, heat kernal signature |
|
3/8
|
Functional Map | functional map |
|
3/13
|
Cycle consistency | map network view of data, enforcing consistency among data by cycle constraint |
|
3/15
|
Deep Learning on Graphs | deep learning on graphs |
|
3/15
|
Assignment 4 | due: Mar 20th, 2018 |