liveBook · Manning

ConvexNet

As explained in Optimization Algorithms: AI techniques for design, planning, and control problems

ConvexNet is the pointer-network model instance described for predicting convex hull point sequences.

ConvexNet predicts the ordered point sequence forming a convex hull. It is a pointer-network model implemented with a Transformer encoder and decoder using multi-head self-attention. section 11.9

Predicting a 50-point hull

For one test sample, ConvexNet predicted [0, 3, 5, 31, 45, 47, 48, 40, 10], exactly matching the true sequence and achieving 100% hull overlap. A comparison plot places the SciPy and ConvexNet hulls side by side.

How Optimization Algorithms: AI techniques for design, planning, and control problems covers it

How deep it goes
Descends through Transformer architecture, hyperparameters, training, validation, testing, visualization, and hull-overlap measurement.
Examples
Uses generated point datasets, predicted index sequences, accuracy and overlap benchmarks, and SciPy-versus-ConvexNet hull plots.
What you can do after
Build it from scratch with PyTorch, train it on generated point sets, and assess predictions using accuracy and hull overlap.
How it connects
Connects pointer networks, Transformers, self-attention, supervised learning, convex hulls, PyTorch, and SciPy.
Cover of Optimization Algorithms: AI techniques for design, planning, and control problems
liveBookmanning.com
Read Optimization Algorithms: AI techniques for design, planning, and control problems on liveBook Get the book at manning.com

This explainer was generated by Manning's knowledge graph from the text of Optimization Algorithms: AI techniques for design, planning, and control problems. The section links point to the parts of the book it draws on.