Topics and Lectures

Topic 0: Overview

Lecture 1: Overview

Lecture 2: Math Preliminaries

Topic 1: Representing Designs
Learning Representations:

Lecture 3: Learning Shape Embeddings

DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation

Vitruvion: A Generative Model of Parametric CAD Sketches

Lecture 4: Learning Templates

Neural Cages for Detail-Preserving 3D Deformations

Joint Learning of 3D Shape Retrieval and Deformation

A Revisit of Shape Editing Techniques: from the Geometric to the Neural Viewpoint

Programs:

Lecture 5: DSL Design Decisions

Penrose: From Mathematical Notation to Beautiful Diagrams

Carpentry Compiler

Taichi

Szalinski

Lecture 6: Automatic DSL design from examples

Data-Efficient Graph Grammar Learning for Molecular Generation

Inverse Procedural Modeling of Branching Structures by Inferring L-Systems

ShapeMOD: Macro Operation Discovery for 3D Shape Programs

DreamCoder: Growing generalizable, interpretable knowledge with wake-sleep Bayesian program learning

Learning Design Patterns with Bayesian Grammar Induction

Topic 2: Instantiating Design Representations and Reverse Engineering to Match Input

Lecture 7: Design Translation

Im2Vec

A Learned Representation for Scalable Vector Graphics

DeepSVG: A Hierarchical Generative Network for Vector Graphics Animation

Supervised Fitting of Geometric Primitives to 3D Point Clouds

Deep Parametric Shape Predictions using Distance Fields

Learning Fuzzy Set Representations of Partial Shapes on Dual Embedding Spaces

Lecture 8: Program Synthesis

Inferring CAD Modeling Sequences Using Zone Graphs

Sketch2CAD

Functional Programming for Compiling and Decompiling Computer-Aided Design

Learning to Infer Graphics Programs from Hand-Drawn Images

InverseCSG: Automatic Conversion of 3D Models to CSG Trees

Lecture 9: Neurosymbolic Synthesis

Learning to Infer and Execute 3d Shape Programs

DeepCAD

Hierarchical Motion Understanding via Motion Programs

Synthesizing Programs for Images using Reinforced Adversarial Learning

ShapeAssembly: Learning to Generate Programs for 3D Shape Structure Synthesis

Topic 3: Performance-Driven Design

Lecture 10: Performance-Driven Design

Accelerated discovery of 3D printing materials using data-driven multiobjective optimization

Diversity-Guided Multi-Objective Bayesian Optimization With Batch Evaluations

AutoOED: Automated Optimal Experiment Design Platform

Fabrication-in-the-Loop Co-Optimization of Surfaces and Styli for Drawing Haptics

Lecture 11: Multi-Objective Optimization

Multi-Task Learning as Multi-Objective Optimization

Exact Pareto Optimal Search for Multi-Task Learning: Touring the Pareto Front

Multi-Objective Optimization by Learning Space Partitions

Efficient Continuous Pareto Exploration in Multi-Task Learning

Lecture 12: Bi-level Optimization

Co-Optimization of Design and Fabrication Plans for Carpentry

RoboGrammar: Graph Grammar for Terrain-Optimized Robot Design

Retrieval on Parametric Shape Collections

An overview of bilevel optimization

Lecture 13: Design Space Exploration

Mixplorer: Scaffolding Design Space Exploration through Genetic Recombination of Multiple Peoples’ Designs to Support Novices’ Creativity

Sequential Line Search for Efficient Visual Design Optimization by Crowds

Sequential Gallery for Interactive Visual Design Optimization

Interactive Optimization of Generative Image Modeling using Sequential Subspace Search and Content-based Guidance

Design Adjectives: A Framework for Interactive Model-Guided Exploration of Parameterized Design Spaces

Lecture 14: More Design Space Exploration

Exploratory Modeling with Collaborative Design Spaces

Designing with interactive example galleries

Parallel prototyping leads to better design results, more divergence, and increased self-efficacy

Data-driven suggestions for creativity support in 3D modeling

DesignScape: Design with Interactive Layout Suggestions

Providing Timely Examples Improves the Quantity and Quality of Generated Ideas

Lecture 15: Constraints

A review on geometric constraint solving

SketchGen: Generating Constrained CAD Sketches

CAD as Language: Conditionally generating CAD programs with constraints

RELATE: Physically Plausible Multi-Object Scene Synthesis Using Structured Latent

Conflict Driven Clause Learning Algorithm (SAT)

A Geometric Constraint Solver

Lecture 16: Hierarchical/Partial Specifications

Interactive Architectural Design with Diverse Solution Exploration

Computational network design from functional specifications

BSP-Net

Learning Hierarchical Shape Segmentation and Labeling from Online Repositories

Computational Design of Hierarchically Structured Materials

Bioinspired hierarchical composite design using machine learning: simulation, additive manufacturing, and experiment

Computational Design of Reconfigurables

Note: Lecture slides will be polished and updated throughout the course.