Dear All,
We are pleased to inform you that our paper has been accepted at
IC4E2011, Mumbai conference.We plan to submit the extended version
for journal publication soon.
We present an abstract of our paper. The proceedings will appear in
IEEE conference proceedings soon.
===================
Law of Connectivity in Machine Learning
Aim: - We present a new law in the field of machine learning. We
believe that there is always a connection present between two
entities. We state that for any scenario, a subject entity is always,
directly or indirectly,connected and affected by single or multiple
independent / dependent entities, and their impact on the subject
entity is dependent on various
factors falling into the following categories:-
1. Existence of entity
2. Inner state of the entity
3. External state of the entity
4. State of communication of the entity.
We can present the above law in the following manner:-
Distance (En) = f(X) = n=0∫∞ {E (n-1) <--->E (n) } > 0
where E (n), E (n-1) > 0 and E is the
entity (with index ‘i’) attributes
(mentioned above) which are involved in the interaction. We postulate that
the required attributes may be added further, depending upon the
requirements of the learning algorithms, besides the one shown above. Thus,
we measure them as Ei in the above equation.
We further derive the equation to be:-
Distance (En) = f(X) = n=0∫∞ (E (n) / E (n-1)) * (E (n-1) / E (n-2)) ……. *
(E (1) / E (0))
Thus, En > 0 and 1 > Ei > 0. A value of 0 defines that the interaction is
not possible due to the entity’s inability to pursue the interaction. Again,
the value can never been 1 as there is always some disturbance or obstacle
that will hinder the interaction. Again, the measurement is probabilistic
and En > 0 and En < 1.
Where En is the intended initiator and E0 is the intended recipient of the
interaction. The entities considered in the interaction depend on the path
that is chosen for learning and considering this interaction. The initia
tor
may a take a mirror and look at the image of the recipient, or talk to
that
person over the phone or even go personally to meet the recipient. Every
path chosen, depending on the learning algorithm, will profoundly impact the
selection of entities and thus decide the quality and time of the
interaction. The number of entities varies in any learning scenario and thus
plays an important role in the quality of interaction.
This law will make the foundation of neural networks, decision tree based
techniques and other learning algorithms. This law will allow the
implementation of long as well as facilitate deeper investigations of
networks in learning models. Since any model is prone to connection with
some other entity, it has to decide whether it wants to connect or not. This
decision decides whether the interaction will happen or not. The absence of
interaction between two entities means that the path to connect the entities
is being ignored or difficult to follow. However, we postulate that the
connection path always exists but the abil
ity of the entities to become
aware and use the connection is what is missing. The latter decides the
scenario and the scope of the interaction too.
Keywords:* Machine Learning; entity; independence; interaction
===================
Please feel free to give us your comments and feedback.
Thanks & Regards,
Jitesh Dundas
--
Thanks & Regards,
Jitesh Dundas
http://openwetware.org/wiki/Jitesh_Dundas_Lab
Phone:- +91-9561527190
" No one is above nature in this world. If anything can be taken from
one human to another, then the same can be given to a machine as well.
The ways of nature of controlling us are very strange" - Jitesh Dundas
“Through dangers untold and hardships unnumbered I have fought my way
here to take back what you have stolen.”
-- Labyrinth (The Movie (1986))
_______________________________________________
uai mailing list
uai@ENGR.ORST.EDU
https://secure.engr.oregonstate.edu/mailman/listinfo/uai
