14 Jan Timing is Everything: Asynchronicity of E/I
Time marches straight on in a never-ending rhythm, never stopping to rest – never missing a beat. After memory fails, and all the mainsprings unwind, time strides forward. Coincidence and synchronicity of events in the physical universe are time bound, but not necessarily meaning bound. In the human brain, time loses relevance and thought flows in lesser and greater impulses, like the economy of a great nation, deriving or inventing meaning to suit the moment. The flow of time is inexorable. The rise and fall of meaning is transient.
Brain Impulses in Time
Today, I’d like to talk a little more about the flow of positive or excitatory and negative or inhibitory electric potentials, collectively E/I, between neurons in the brain. The arrangements of links between neurons and the specialization of areas in the brain make the human cognitive apparatus very complex. Both of these factors are spatial (sometimes referred to as structural). The locations of E/I activity within this space provide important keys to understanding our ability to think. The spatial aspects of our cognitive physiology, however, are complemented by equally complex processes that are temporal in nature. Besides affecting the intensity or weight of action potential transmission, many of the factors mentioned in the constraints on E/I flow (such as concentrations of sodium, potassium, and neurotransmitter) can affect the duration of impulses. This introduces an important temporal element to an otherwise spatially oriented architecture.
Weights or intensities of impulses can be described as falling within the spatial context because they can be modeled in the structure of a spatially oriented system architecture, such as a typical neural network. Temporal factors, however, require functional algorithms or heuristics for those temporal functions that will be applied to the information in the architecture.
| Understanding Context Cross-Reference |
|---|
| Click on these Links to other posts and glossary/bibliography references |
|
|
 |
|
| Prior Post | Next Post |
| Visual Input Processing | The Chromophore as Digital Bit |
| Definitions | References |
| heuristics E/I links | Wikipedia and Carl Jung |
| think context | Coincidence |
| neural network | Jim Fournier on synchronicity |
In simpler terms… If all impulses in the human brain were 1-millisecond spikes, designing a simulator would be simple. Since the time wouldn’t change between impulses, the time factor would not need to be modeled. The scheduling would be simple and the architecture of the processing space would be the most critical thing. Since activation does have variable time properties, any accurate process scheduling model has to consider the natural asynchronous flow. A comprehensive simulation would need algorithms to model the asynchronous aspects of the flow of excitation and inhibition in the processing apparatus, hence, timing is everything. 
Real World Models
In the real world, as well as in computing, synchronicity is an effective mechanism of binding events and communications together. But even disjoint events and communications have important places. Take, for example, SMS text messaging or texting. When people are speaking synchronously in person, or remotely over the phone, there is an immediacy that facilitates understanding. But sometimes, immediacy is not possible, and sometimes it is even the enemy of understanding. I’m not going into that now, but I propose that both synchronous and asynchronous neural models in artificial systems play important roles.
Asynchronous Transmission
Understanding the impact of the interaction between local and action potentials is important in interpreting the temporal element of spreading activation in the brain. Add to that the electro-chemical processes mediating decay, and it appears that nerve cells exhibit little synchrony: any given input may be influencing cells in different parts of the brain long after the stimulus has passed. With successive stimuli, an input received at time T may still be the subject of cognitive activity while input T + 10 is being processed. The asynchronous and acyclic elements of the brain’s functionality are, perhaps, some of the most complex attributes to model either in computer hardware or in mathematical algorithms. These functions call for fuzzy algorithms that support either of these two approaches:
- non-deterministic techniques for simulating chaotic, non-directional flow of activation in a large-scale neural network that lead to a desired outcome, or
- huge numbers of constraints that activate complex threads of weighted reasoning in a large knowledge network that lead to a desired outcome.
Parallel programming languages such as ADA have built-in mechanisms for many simultaneous processes to operate independently then come together to deliver an outcome. Information-Space based models also provide mechanisms for an arbitrary number of disparate processes to independently affect an outcome. Most neural networks use time-sequenced directional activation that does not reflect the natural order. In an asynchronous network, flow of activation is chaotic. Distributed systems like this, and, for example, a free-market economy, are great models of efficient and robust processes needed to achieve complex outcomes.
The level of parallelism differs by many orders of magnitude between parallel programs and the flow of impulses in the human brain, but the models needed to produce valid outcomes may be remarkably similar. I have found this type of Space-Based Model very effective in a variety of computing tasks and process architecture patterns.
| Click below to look in each Understanding Context section |
|---|
SHARE THIS ARTICLE
