Turn abandoned patents
into your own filing
into your own filing
This describes a way to automatically build a neural network that performs well. It starts with a basic network and repeatedly modifies it by trying out different changes from a predefined set of options. Each modified network is then trained, and if it meets certain performance goals, it becomes the final, optimized network; otherwise, the process continues with further modifications until a suitable network is found.
Why it matters: Filed before advanced AutoML and Neural Architecture Search became mainstream. The method's iterative network design is now more efficient and relevant given current AI model complexity and computational resources.
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
