Solving mixed integer programs using neural networks

This invention describes a way to use artificial intelligence to help solve complex math problems called Mixed Integer Programs (MIPs). It works by taking the problem's details and feeding them into a special type of AI called an encoder neural network. This network then creates unique digital representations, or "embeddings," for the parts of the problem that must be whole numbers. These embeddings are used to suggest possible partial solutions, which are then expanded into full candidate solutions, and the best one is chosen. The method specifically focuses on using these embeddings to guide the assignment of values to a subset of the integer variables.

Why it matters: Neural network architectures and training techniques have advanced significantly since 2021, potentially making the generation of effective embeddings and guided constraint formulation more robust and efficient for complex optimization problems.

Status: Under examinationA pending application. It has not been granted, and it has not been abandoned.
How hard to build: SpecializedMachine learning software, optimization algorithms

3 AI reinvention directions for this patent

See all three

AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.

Reinvent this with AI

Where this one came from

Every morning we go through millions of abandoned patents and pick out 15 worth your time. This was one of today’s.

This one

+10