Ofer Matan

Learn More
We present a feed-forward network architecture for recognizing an uncon-strained handwritten multi-digit string. This is an extension of previous work on recognizing isolated digits. In this architecture a single digit rec-ognizer is replicated over the input. The output layer of the network is coupled to a Viterbi alignment module that chooses the best(More)
A model for on-site learning is presented. The system learns by querying \hard" patterns while classifying \easy" ones. This model is related to query-based ltering methods, but takes into account that in addition to labelling, ltering through the data has a cost. A few simple policies are introduced and analyzed for a simple problem (1D high low game). In(More)
A neural network algorithm-based system that reads handwritten ZIP codes appearing on real US mail is described. The system uses a recognition-based segmenter, that is a hybrid of connected-components analysis (CCA), vertical cuts, and a neural network recognizer. Connected components that are single digits are handled by CCA. CCs that are combined or(More)
We study the classiication ability of majority-vote ensembles of classiiers. A majority ensemble classiies a pattern by letting each member of the ensemble cast a single vote for the correct class and decides according to a simple majority or a special majority vote. We give upper and lower bounds on the classiication performance of a majority ensemble as a(More)
The staff scheduling problem is a critical problem in the call center (or more generally, customer contact center) industry. This paper describes Director, a staff scheduling system for contact centers, Director is a constraint-based system that uses AI search techniques to generate schedules that satisfy and optimize a wide range of constraints and service(More)
■ The staff scheduling problem is a critical problem in the call center (or, more generally, customer contact center) industry. This article describes DIRECTOR , a staff scheduling system for contact centers. DIRECTOR is a constraint-based system that uses AI search techniques to generate schedules that satisfy and optimize a wide range of constraints and(More)
Researchers have long realized that supervised learning of a task by training on data drawn from the target distribution and labeled by an expert may be tedious and ineecient. Much of the data in training sets may be redundant and a large number of examples may be needed to gain reliable sampling of the distribution tails. The costs associated with(More)
We study Schapire's Boosting Algorithm(SBA) for use in practice. SBA is analyzed in terms of its representation and its search. We show that the SBA representation is a piecewise tiling of the domain and that if the weak learner has low coverage ability, SBA's search may fail to boost or may give a sub-optimal solution. We present a rejection boosting(More)
  • 1