Ofer%20Dekel,%20Shai%20Shalev-Shwartz,%20Yoram%20Singer

1
Smooth e-Insensitive Regression by Loss
Symmetrization

• Ofer Dekel, Shai Shalev-Shwartz, Yoram Singer
• School of Computer Science and Engineering
• The Hebrew University
• oferd,shais,singer_at_cs.huji.ac.il
• COLT 2003 The Sixteenth Annual Conference on
  Learning Theory

2
Before We Begin
Linear Regression given find
such that Least Squares minimize Support
Vector Regression minimize s.t.
3
Loss Symmetrization
Loss functions used in classification Boosting
Symmetric versions of these losses can be used
for regression
4
A General Reduction

• Begin with a regression training set
• where ,
• Generate 2m classification training examples of
  dimension n1
• Learn while maintaining
• by minimizing a margin-based classification loss
  

5
A Batch Algorithm

• An illustration of a single batch iteration
• Simplifying assumptions (just for the demo)
• Instances are in
• Set
• Use the Symmetric Log-loss

6
A Batch Algorithm
Calculate discrepancies and weights
43210
0 1 2 3
4
7
A Batch Algorithm
Cumulative weights
0 1 2 3
4
8
Two Batch Algorithms
Update the regressor
43210
Log-Additive update
0 1 2 3
4
9
Progress Bounds
Theorem (Log-Additive update) Theorem
(Additive update) Lemma Both bounds are
non-negative and equal zero only at the optimum
10
Boosting Regularization

• A new form of regularization for regression and
  classification Boosting
• Can be implemented by addingpseudo-examples
• Communicated by Rob Schapire

where
11
Regularization Contd.

• Regularization ? Compactness of the feasible set
  for
• Regularization ? A unique attainable optimizer of
  the loss function

?
Proof of Convergence
Progress compactness uniqueness asymptotic
convergence to the optimum
12
Exp-loss vs. Log-loss

• Two synthetic datasets

Log-loss Exp-loss
13
Extensions

• Parallel vs. Sequential updates
• Parallel - update all elements of in parallel
• Sequential - update the weight of a single weak
  regressor on each round (like classic boosting)
• Another loss function the Combined Loss

Log-loss
Exp-loss
Comb-loss
14
On-line Algorithms

• GD and EG online algorithms for Log-loss
• Relative loss bounds

Future Directions

• Regression tree learning
• Solving one-class and various ranking problems
  using similar constructions
• Regression generalization bounds based on natural
  regularization