Introduction

A device to simultaneously compute several inputs and build a classification process. The invention exploits physical processes that occur in response to external stimuli within complex nanostructured materials and it is an improved alternative to classic artificial neural networks.

Technical features

A device to implement a multi-input, re-configurable, Boolean classifier that can recognise several binary inputs. The classifier can be reprogrammed without the employment of learning algorithms, thus with a reduction of time and energy costs. Different from classic artificial neural networks, the innovative device does not require a detailed design of its components because it exploits self-assembled nanostructured films and their physical properties. This peculiar feature makes the system more tolerant to errors occurring during the production process and the functioning. Our innovative method set the basis for developing new devices able to compute big data to accomplish tasks such as “pattern recognition”.

Possible Applications

• Images recognition;
• Robotics;
• Data compression;
• Cryptography and data encryption.

Advantages

• Self-assembled materials, simple and cheap production;
• No need of learning processes thus more efficient;
• Higher tolerance to errors.