Fwd: Recursive Tree in Python

[Fred Kitoogo]

Note: forwarded message attached.


Kitoogo Fredrick Edward


        
        
                
___________________________________________________________ 
All new Yahoo! Mail "The new Interface is stunning in its simplicity and ease 
of use." - PC Magazine 
http://uk.docs.yahoo.com/nowyoucan.html

--- Begin Message ---

Hi there,   I wish to create a recursive tree and test in Python for combining classifiers, below are the two Algorithms (Training & Testing):-   Recursive Stacking Algorithm for Training Input: Training Set S Base learning algorithms A1, ...,At Base Classifiers for combination Ct = {c1,.....,ct} Cross-Validation partition fold J (2 <J<_ n) Conflict ratio threshold  T Output: A Recursive-Stacking tree _____________________________________________________________________________________________________ begin 1. Call stacking to generate meta-level training set Mtrain 2. Analyze Mtrain for conflict types (Call recursive stacking inside if stopping criteria is not met) • for all (c1, c2, ..., ct) conflict ratio e<=T, mark it as a terminator conflict • for all (c1, c2, ..., ct) conflict ratio e>T  if stopping criteria is not met then  begin take all base level training data of type (c1, c2, ..., ct) into S(c1,c2,...,ct) call Recursive Stacking-Training on S(c1,c2,...,ct) end else mark this conflict type as a terminator conflict 3. Return the Recursive Stacking tree end __________________________________________________________________________________________________________   Recursive Stacking Algorithm for Testing Input: Testing Data: D Output: Classifications   ______________________________________________________________________________________________________________ begin 1. for all (xi, ?) E D call stacking to return meta-level testing item (C1(x), ....,Ct(x)) and put into Mtest 2. for all (xi, ?) E D put it into subset D(c1,c2....,ct) according to its type (c1, c2...., ct) in Mtest 3. for each subset D(c1,c2....,ct) if (c1, c2...., ct) is a terminator conflict then for all (xi, ?) 2E D(c1,c2,...,ct) use the majority label of type (c1, c2, ...., ct) else call recursive stacking - Testing with D(c1,c2,...,ct) and use the returned classifications 4. Combine predictions for each subset and return end _________________________________________________________________________________________________________________   Can somebody please help?     Thanks,     Kitoogo Fredrick Edward PHD Student email: [EMAIL PROTECTED]           [EMAIL PROTECTED]           [EMAIL PROTECTED] Tel:     256-41-233423           256-77-2-855884

--- End Message ---

-- 
http://mail.python.org/mailman/listinfo/python-list