Knowm Inc was founded to develop memristive machine learning hardware and promote memristor science.

Knowm Memristors

The Knowm memristor material stack is based on mobile metal ion conduction through a chalcogenide material that has undergone a metal-catalyzed chemical reaction that creates channels which constrain the flow of metal ions. The resistance is related to the amount of metal located within the active channel layer. Doping materials in the active layer enhance and optimize properties such as switching speed, switching energy, endurance, data retention, and incremental sensitivity. Knowm Memristors are available for sale and are shipping worldwide.

AHaH Computing

Anti-Hebbian and Hebbian (AHaH) computing is a new form of computing based on collections of differential-pair memristor synapses. It has been been shown to offer general solutions to memory, reconfigurable logic and machine learning.

Thermodynamic RAM (kT-RAM)

kT-RAM provides a universal synaptic integration and adaptation substrate for AHaH Computing. The kT-RAM architecture allows for drastic reductions of energy associated with machine learning operations by reducing synaptic adaptation and inference to local analog operations and eliminating the need to communicate and convert analog information by reducing communication to sparse spike (integer) streams that directly index kT-RAM memory addresses.

Knowm API

The Knowm API is a software hook to kT-RAM, where machine learning functions have been reduced to kT-RAM instruction set routines. 

Knowm Technology Stack

The kT-RAM Technology Stack is a specification that goes from memristors to machine learning applications. This allows separate groups to specialize at one or more levels of the stack where their strengths and interests align. Improvements at various levels can propagate throughout the whole technology ecosystem, from materials to markets, without any single participant having to bridge the whole stack.

More Information