anytime algorithms for learning anytime classifiers saher ... - Technion

Technion - Computer Science Department - Ph.D. Thesis PHD-2008-12 - 2008
a1
a1
a1
a1
a1
a1
a1
E-MC=
$110
a1
a1
E-MC=
$250
a1
a1
E-MC=
$170
a1
ρ c = $60
Figure 5.3: Attribute evaluation in Pre-Contract-TATA. E-MC stands for the expected
misclassification cost. For each candidate split, we sample the space of trees
under it that fit the remaining budget ($60 in the example) and evaluate the split by
the minimal expected misclassification cost in the sample ($110 in the example).
linearly with r, just as it does in ACT (Esmeir & Markovitch, 2007a). When we
cannot afford sampling (r = 0), TATA builds the tree using C4.5$.
5.1.4 Interruptible Learning of Pre-contract Classifiers
The algorithm presented in Section 5.1.3 requires r, the sample size, as a parameter.
When the learning resources are not predetermined, we would like the
learner to utilize extra time until interrupted. In Chapter 3 we presented IIDT,
a general framework for Interruptible Induction of Decision Trees, that need not
be allocated resources ahead of time. IIDT starts with building a greedy tree.
Then, it repeatedly selects a subtree whose reconstruction is expected to yield
the highest marginal utility, and rebuilds the subtree with a doubled allocation
of resources.
The same iterative improvement approach can be applied to convert preconstract-TATA
into an interruptible algorithm. The initial greedy tree would be
built with C4.5$, and subtree reconstructions would be made using pre-contract-
TATA. The marginal utility of constructing a tree would take into account both
the expected misclassification cost of the tree and the expected resources required
104
a1
a1
a1
a1
The space of trees whose test cost ≤ $60