Current research: neural networks with temporal bais and governing function.

ordinary neuron network: signals has temporal bais based on layers and circuit design.
ordinary learning function: characterising signal magnitude.
temporal bias archived by computation: simple event driven OO programming.
circuit realization: integration operator (recurrent nn).

Main-point :
PoV A. Learning function characterising “signal time shifts” and chronicle events.
PoV B. Suppressing error / Provisioning implied trends , without delaying learning / adaptation

ripple

X is the matter of Subject.

It is governed / disturb by lots of events. Some of the events are predictable and thus to be considered as outstanding sources to for learning. The effect (from) of these events are some how understood, but the triggering time have to be studied / due to change.

Successful learning provides faster and more precise adaptation, outlining temporal relation of centers of decision.

Traditional model:
Error -> tune signal magnitude -> leads to signal decay -> lost track

Proposed model:
Error -> delay / stretch signal -> implies stacks ->
A. stack limits / overflow ?
B. lost in choosing stack / trapped ?
C. better responce to 1st wave ?
D. unefficient learning ?

B/D -> implies second layer of learning/decision (speedy PCA)
Objective : Optimizing C/A

Pros:
- Real life analysts face the same problem set. There is hope to enhance these projections by computation.

Cons:
- Traditional model may be enough, it does not consider human concept / modelling. New model worths no more than satisfying error-poned “model / explanation” needs.