Miraclebet

2013-05-19T15:34:11Z

46.7.236.155: rm nonsense

{{multiple issues|orphan =August 2010|COI=December 2009|notability =December 2009|primarysources=December 2009}}

'''Regulatory Feedback Networks''' describe a class of neural networks
related to '''Virtual Lateral Inhibition''' (named to distinguish it
from [[lateral inhibition]]) that perform inference using [[negative
feedback]].<ref>J. Reggia, “Virtual lateral inhibition in parallel
activation models of associative memory,” in Proc. 9th International
Joint Conference on Artificial Intelligence., Aug. 1985, pp.
244-248.</ref><ref name=mcfadden>Mcfadden, F. E. (1995).
"Convergence of Competitive Activation Models Based on Virtual Lateral
Inhibition." Neural Networks 8(6): 865-875.</ref><ref
name=first>Achler, T. (2002). Input Shunt Networks. Neurocomputing,
44, 249-255.</ref> The feedback is implemented during recognition
and during recognition connectivity parameters are not changed. Thus
this is completely separate from learning/training (e.g. [[supervised
learning]] or [[unsupervised learning]]). This is also different from
models of [[attentional shift|spatial attention]]. Instead, these
networks determine the relevance of inputs through a "conservation of
information principle".

== How the network functions ==
The computational basis of conservation of information is that an input
should not pass more information than is justified to the next layer.
Thus inputs are regulated by the outputs they activate. Subsequently,
each input’s contribution (i.e.
[http://www.scholarpedia.org/article/Visual_salience salience]) is
adjusted through feedback regulation by its associated outputs. The
amplitudes of the adjusted inputs are propagated to the output layer. A
new salience is re-evaluated based on the new output activity (through
feedback). This can be iterated until the networks reach steady
state.<ref name=mcfadden /> At every step, the role of salience is
to maintain the relation where: the total activity of outputs connected
to an input will be equivalent to the input’s amplitude.<ref
name=first /><ref name=agi>Achler T., Amir E., “Input Feedback
Networks: Classification and Inference Based on Network Structure”
Artificial General Intelligence 2008
[http://reason.cs.uiuc.edu/tsvi/AGI.pdf pdf]</ref>

== How the network is used ==
These networks are best suited for nodes with binary connections.<ref
name=first /><ref name=agi /><ref name=shedding>Achler
T., Omar C., Amir E., “Shedding Weights: More With Less”, IEEE Proc.
International Joint Conference on Neural Networks, 2008
[http://reason.cs.uiuc.edu/tsvi/IJCNN.pdf pdf]</ref> Instead of
weights determining the relevance of connections, input salience is
adjusted at the time of recognition. For example, a node representing
car may connect to features wheels, door, and bumper. A node
representing bicycle may connect to features wheels, pedals, and chain.
Given wheels, the network will determine how relevant the wheels are to
either the bicycle or car nodes during recognition.

== Benefits/costs ==
This model displays unparalleled performance given simultaneous
patterns, addressing [[Curse of dimensionality|combinatorial
explosions]] associated with simultaneous patterns.<ref
name=shedding /><ref>Achler T., Vural C., Amir, E., "Counting
with Biologically Inspired Regulatory Feedback Networks”, IEEE Proc.
International Joint Conference on Neural Networks, 2009
[http://reason.cs.uiuc.edu/tsvi/counting.pdf pdf]</ref>

The model can also generate solutions composed of multiple output nodes
with minimal overlap.<ref name=agi /><ref>Achler T., “Using
Non-Oscillatory Dynamics to Disambiguate Simultaneous Patterns”, IEEE
Proc. International Joint Conference on Neural Networks, 2009
[http://reason.cs.uiuc.edu/tsvi/Dynamics%20for%20Disambiguation%20IJCNN%202009.pdf pdf]</ref> This property groups patterns together in a manner
that suggests a way out of a fundamental recognition conundrum called
the [[binding problem]] ('unity of perception' version).

In contrast to conventional neural networks or machine learning methods
these networks cannot be guaranteed to be able to capture any arbitrary
pattern. However for the patterns they can capture, they show these
properties.

== Implementation ==
Suppose there are [[Fuzzy logic|fuzzy-type]] input features
<math>x</math> and output nodes <math>y</math>.
Each output node <math>y_j</math> is defined by set of
feedforward binary connections <math>FF_j</math> from
<math>x</math>'s. It also has a set of symmetrical feedback
connections <math>FB</math> that implement [[negative
feedback]]. Due to the symmetry each member of
<math>FF_j</math> (a connection from input to output) has a
corresponding member in <math>FB</math> (a connection from
the same output to same input) that returns and inhibits the input.
Lets label <math>FB_i</math> the set of connections that
return to an <math>x_i</math>.
<math>|FF_j|</math> is the number of connections to
<math>y_j</math>. Lets label <math>s_i</math>
the salience of input <math>x_i</math>. Then the activity of
the output node is determined by:

<math>y_j(t+\Delta t)=\frac{y_j(t)}{|FF_j|}\sum_{k\in
FF_j}s_k</math>.
<br/>
The salience value <math>s_i</math> of a given
<math>x_i</math> is determined by:
<br/>
<math>s_i =\frac{x_i}{\sum_{r\in{FB_i}}y_r(t)}</math>.
<br/>
These equations can be iterated until the network reaches steady state.

== See also ==
* [[Visual perception]]
* [[Visual Object Recognition in Cognitive Neuroscience]]
* [[Bag of words model in computer vision]]
* [[Computational neuroscience]]

== References ==
<references/>

{{DEFAULTSORT:Regulatory Feedback Network}}
[[Category:Control theory]]
[[Category:Computational neuroscience]]

formulasearchengine - User contributions [en]

Miraclebet