Our Approach

The human brain is comprised of over 80 billion neurons, all communicating with one another via electrical pulses. These signals are commonly referred to as action potentials or spikes, and their combined activity brings rise to all of our cognitive function, movement, and perception. In order to suppress the unwanted effects of various neurological disorders, we require a means to intelligently manipulate these neural signals as desired. However, the current state-of-the-art methods are simplistic in nature and limited in what they can achieve. As such, we leverage the following five properties in our algorithm design, to better equip implantable brain computer interfaces (BCI) and neural stimulators for the next era of neural control.

Nonlinear

Neural algorithms take in input signals (neural recordings) and spit out an output signal (stimulation). As our algorithms are nonlinear, we able to learn complex input-output
nonGaussian

Non-Gaussian

Many algorithms assume that the world is Gaussian, which is loosely synonymous with standard. Non-gaussian means that our algorithm can work in non-standard settings.

Non-Episodic

Episodic systems have a natural end state. For example, losing and respawning in a video game. In contrast, non-episodic systems, such as the brain, have no obvious end state, meaning that we can’t simply reset the system over and over again.

Non-Markovian

The system’s past has a long-lasting effect on how it evolves over time. The next state of the system is not just dependent on where it is in the present moment.

Non-Stationary

When the rules of how the system behaves change at one or more points in time. As an example, ocean currents are non-stationary as current patterns are constantly changing.

Thank You!




We've recieved your info and look forward to chatting!