Interpreter is a superstructure that is built on
associative self-building neural network. It combines tick data feed and
neural network to provide real time and intraday live analysis and
foretells. Artificial Foreteller activates Interpreter only when it
predicts future trade period. Activation process is the following -
Preprocessor examines tick data and generates signal that current trade
period is over. That means trend have already changed its direction. Preprocessor
transmits calculated values of different criteria to Network Builder that
activates corresponding receptors in receptor layer. Associative
self-building network propagates activating signals from receptor layer
upward. Each network node can be in one of the following state:
- Node is activated if it
gets activated signals from all input connections (These type of nodes and connections are marked with red
color on Figure below).
- Node is partially activated
if at least one of its input connections does not bring activated
signal (These type of nodes and connections
are marked with rose color on Figure below).
- Node is not activated if
all its input connections do not bring activated signal (These type of nodes and connections are marked with blue
color on Figure below).
If similar trade periods happened before and they
were used for system train, then part of the network that represents these
periods will be activated. Interpreter compares current situation with
those that were used for train and makes conclusion for the future trade
period.
As it can be seen from the figure, only a small
fragment of associative self-organazed neural network is activated. This
feature of associative neural network greatly reduces scope of data that
interpreter should process to make conclusion. In the end of activation
process interpreter receives activated signals from upper layer of the
network and analyses activated and partially activated segments of neural
network. Result of interpreter work is a conclusion whether next trade
period is favorable for trade or not.
|
