Turn abandoned patents
into your own filing
into your own filing
This invention describes a tiny electronic component with two connection points that can change how easily electricity flows through it in a non-linear way. This change in electrical resistance happens when specific signals are sent to its terminals. The component is designed to act like a memory and a processor for artificial intelligence, specifically for training neural networks, by adjusting its resistance to represent "weights" in the network. The signals used to control it can be random sequences of pulses or changes in signal strength.
Why it matters: Filed before the explosion in deep learning model sizes and the demand for energy-efficient AI hardware. The need for specialized in-memory computing solutions to accelerate neural network training has intensified significantly since 2015, making novel RPU architectures more relevant.
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
