Core Terms

The words NuMorph uses — ERN-AI, ALB, NESA — connect to the terms the market actually searches: low-power AI, edge inference, event-driven vision. This glossary bridges the two.

Neuromorphic computing

Computing architectures modeled on the structure and behavior of biological nervous systems, typically characterized by event-driven processing, sparse activation, and very low power consumption relative to conventional von Neumann architectures.

Event-driven vision

A sensing paradigm in which visual data is generated only when the scene changes, rather than captured at a fixed frame rate. Data volume scales with scene activity instead of with time.

Frame-based vision

The conventional imaging paradigm in which a sensor captures complete images at a fixed rate regardless of scene content, producing substantial redundancy in static or slow-changing scenes.

Analog AI

AI computation performed using analog physical quantities rather than discrete digital arithmetic, typically achieving significantly lower energy per operation for suitable workloads.

Edge inference

Running AI model inference on or near the sensor rather than transmitting data to a centralized server or cloud, reducing latency and bandwidth requirements.

Perception-native AI

AI architecture in which perception processing is structured around the sensing modality itself rather than adapting general-purpose models to sensor output after the fact.

Physical AI

AI systems that perceive and act in the physical world, including robots, drones, autonomous vehicles, and industrial automation systems.

Spiking neural network (SNN)

A neural network model in which neurons communicate through discrete events (spikes) over time rather than continuous activation values, closely mirroring biological neural signaling.

ERN-AI (Evolutionary, Reconfigurable, Neuromorphic-AI)

NuMorph AI's overarching architecture paradigm: neuromorphic AI systems that are evolvable and reconfigurable in real time, built from the ALB and NESA components.

ALB (AI-Lego-Block)

NuMorph AI's atomic neuromorphic processing unit, derived from honeybee visual neurology, composable into arrays.

NESA (Neuromorphic Edge Sensing Array)

NuMorph AI's vision sensing array architecture, organizing ALB arrays into a layered structure that is reconfigurable in real time.

SWaP (Size, Weight, and Power)

A combined design metric used to evaluate hardware suitability for edge and embedded deployment, where minimizing physical footprint, mass, and energy consumption is critical.

LLM / LVM (Large Language Model / Large Vision Model)

Large-scale neural models trained on text (LLM) or visual data (LVM) to perform general-purpose language or vision tasks.

DNN / CNN / RNN

Common deep learning architectures: DNN (Deep Neural Network) is the general multi-layer form, CNN (Convolutional Neural Network) specializes in spatial data such as images, and RNN (Recurrent Neural Network) handles sequential data through recurrent connections.

FPGA (Field-Programmable Gate Array)

A reconfigurable integrated circuit whose hardware logic can be reprogrammed after manufacturing, often used for prototyping and low-latency inference.

ASIC (Application-Specific Integrated Circuit)

A chip custom-designed for a single application, offering higher efficiency and performance than general-purpose hardware at the cost of flexibility.

SLAM (Simultaneous Localization and Mapping)

A technique enabling a device to build a map of an unknown environment while simultaneously tracking its own position within it.

GPS (Global Positioning System)

A satellite-based navigation system providing location and time information anywhere with an unobstructed line of sight to the satellites.

ROS (Robot Operating System)

An open-source middleware framework providing tools and libraries for building and controlling robotic systems.

RADAR (Radio Detection And Ranging)

A sensing method using radio waves to detect the distance, speed, and direction of objects, commonly used in automotive and physical-AI systems.