AIRIS Labs AI KNN IP

Sample Size

3 options for selection: Programmable learning data size is up to 1024 for 1k option, 2048 for 2k and 4096 for 4k option.

Attribute Size

Up to maximum of 1024 attributes per sample for learning. Programmable attribute size in multiples of 64.

Programmable K

Programmable k value of K-NN from 1 to 15.

Distance Calculation

AIRIS Labs AI KNN support Euclidean, Manhattan and Hamming distance calculation.

Connectivity

Utilize high speed communication of direct writing. Optional SPI interface

Silicon Proven

Silicon proven on TSMC 40LP. Achieves 0.7 ms classification of a 1024 learning data samples of 1024 attributes per learning data sample at 100 MHz clock

Silicon proven on TSMC 40LP.
Enclosed in a 88-pin, 12mm x 12mm QFN package.

AIRIS Labs AI KNN offers two modes for learning data: SPI mode and high speed mode. Through the high speed connection, the users can build different types of AI application with the heavy lifting of AI undertaken by AI KNN chip.

Through the SPI connection, the Cerdik Development Board can be slotted into any board with SPI, such as Arduino Uno, Raspberry Pi and etc, to build a simple yet powerful AI system.