Knowm

The Knowm™ concept is simple, and relies on a force experienced by particles suspended in a solution and exposed to time-varying electric fields. The dielectrophoretic force can be used to attract and repel particles to an electrode gap. The electrical resistance across the gap is dependent on the number of nanoparticles bridging the gap: more particles, less resistance and visa versa. At a basic level, this particle bridge can be thought of as a synapse.

It is known that simple plasticity rules can be used to extract the Independent Components (IC) of a data set. IC's can be thought of as a mechanism for the efficient representation of a data set. A front-end nano-scale processor capable of Independent Component extraction would make possible the placement of inexpensive and sophisticated object recognition capabilities in portable electronics.

To date, assembly and fault tolerance represent the most difficult hurdles to building nanoscale electronics. The same plasticity rules capable of extracting IC's also have the desirable property of active fault tolerance. The fixed points of the plasticity rule provide a dynamic system capable of reconfiguration of synapses so as to adapt to changing environmental conditions and correct internal faults. Neural circuits and synapses can fail and input statistics can change. The same plasticity rules capable of Independent Component Analysis will reconfigure the synapses so as to repair the network.

With Knowm™ technology, a nano-scale network can be built that self-assembles and self-repairs while extracting Independent Components from massive data streams, thereby allowing sophisticated pattern recognition technology in a small package.

The idea is straight forward, and considerable experimental and theoretical evidence show that such a network is not only feasible, but that it's also attainable with current technology in a relatively short time frame. Stated simply, pre- and post-synaptic electrode pulses create time-varying electric fields, which can be used to attract particles to the electrode gaps. The resistance across the electrode gap is a function of particle aggregation at the gap. The connection resistance can thus be controlled by pre- and post-synaptic activity. The result is a nano-scale connection capable of emulating plasticity rules we know to be computationally useful and inherently fault tolerant. The algorithm is the architecture, and the architecture builds and repairs itself.

Introductory Flash animation [~3Mb, high bandwidth rec.]

AHAH Plasticity Flash Demo

Synaptic States Flash Demo

[Why is my computer running slow, why can't I see the title or navigation bar, and why do those particles run every single time I load the page?]