The standard approach to decision tree induction is a top-down, greedy algorithm that makes locally optimal, irrevocable decisions at each node of a tree. In this paper, we study an alternative approach, in which the algorithms use limited lookahead to decide what test to use at a node. We systematically compare, using a very large number of decision trees, the quality of decision trees induced by the greedy approach to that of trees induced using lookahead. The main results of our experiments are: (i) the greedy approach produces trees that are just as accurate as trees produced with the much more expensive lookahead step; and (ii) decision tree induction exhibits pathology, in the sense that lookahead can produce trees that are both larger and less accurate than trees produced without it. 1. Introduction The standard algorithm for constructing decision trees from a set of examples is greedy induction --- a tree is induced top-down with locally optimal choices made at each node, with...

The majority of the existing algorithms for learning decision trees are greedy-a tree is induced top-down, making locally optimal decisions at each node. In most cases, however, the constructed tree is not globally optimal. Furthermore, the greedy algorithms require a fixed amount of time and are not able to generate a better tree if additional time is available. To overcome this problem, we present two lookahead-based algorithms for anytime induction of decision trees, thus allowing tradeoff between tree quality and learning time. The first one is depth-k lookahead, where a larger time allocation permits larger k. The second algorithm uses a novel strategy for evaluating candidate splits; a stochastic version of ID3 is repeatedly invoked to estimate the size of the tree in which each split results, and the one that minimizes the expected size is preferred. Experimental results indicate that for several hard concepts, our proposed approach exhibits good anytime behavior and yields significantly better decision trees when more time is available.

The majority of existing algorithms for learning decision trees are greedy—a tree is induced topdown, making locally optimal decisions at each node. In most cases, however, the constructed tree is not globally optimal. Even the few non-greedy learners cannot learn good trees when the concept is difficult. Furthermore, they require a fixed amount of time and are not able to generate a better tree if additional time is available. We introduce a framework for anytime induction of decision trees that overcomes these problems by trading computation speed for better tree quality. Our proposed family of algorithms employs a novel strategy for evaluating candidate splits. A biased sampling of the space of consistent trees rooted at an attribute is used to estimate the size of the minimal tree under that attribute, and an attribute with the smallest expected tree is selected. We present two types of anytime induction algorithms: a contract algorithm that determines the sample size on the basis of a pre-given allocation of time, and an interruptible algorithm that starts with a greedy tree and continuously improves subtrees by additional sampling. Experimental results indicate that, for several hard concepts, our proposed approach exhibits good anytime behavior and yields significantly better decision trees when more time is available.

Abstract—This paper presents a multiobjective approach for the design of electrical distribution networks. The objectives are defined as a monetary cost index (including installation cost and energy losses cost) and a system failure index. The true Pareto-optimal solutions are found with a multiobjective genetic algorithm that employs an efficient variable encoding scheme and some problem-specific mutation and crossover operators. Results based on 21- and 100-bus systems are presented. The information gained from the Pareto-optimal solution set is shown to be useful for the decision-making stage of distribution network evolution planning. Index Terms—Decision-making, energy distribution networks, genetic algorithms (GAs), multiobjective optimization, network topology optimization. I.

Introduction Top-down induction of decision trees is a simple and powerful method of inferring classication rules from a set of labeled examples 1 . Each node of the tree implements a decision rule that splits the examples into two or more partitions. New nodes are created to handle each of the partitions and a node is considered terminal or a leaf node based on a stopping criteria. This standard approach to decision tree construction thus corresponds to a top-down greedy algorithm that makes locally optimal decisions at each node. There are two advantages that decision trees have over many other methods of classication methods. The rst is that the sequence of decisions made from the root node to the eventual labeling of a test input is easy to follow. This gives them an intuitive appeal that other methods of classication such as

A Robotic System is being developed to automate the crucible packing process in the CZ semiconductor wafer production. It requires the delicate manipulation and packing of highly irregular shaped polycrystalline silicon nuggets, into a fragile glass crucible. Here an on-line algorithm is presented to plan the packing. It uses a method called Virtual Trial and Error. The on-line algorithm handles large numbers of highly irregularly shaped object of different sizes without requiring the object models. Working with the 3-D range maps of objects, it is computationally fast enough to be applied in real-time to practical industrial applications, such as the CZ wafer manufacture. Simulation results show that it compares well with the human performance. The integrated system is shown to achieve high production rates, required precision and cost effectiveness. 1 Introduction During the widely used CZ semiconductor production process, highly irregular shaped polycrystalline silicon nuggets are...

A Robotic System is being developed to automate the crucible packing in the CZ semiconductor wafer production. It requires the delicate manipulation and packing of highly irregular shaped polycrystalline silicon nuggets, into a fragile glass crucible. Here an on-line algorithm is presented to plan the packing. It uses a method called Virtual Trial and Error. The on-line algorithm handles large numbers of highly irregularly shaped object of different sizes without requiring the object models. Working with raw vision data it is computationally fast enough to be applied in real-time to practical industrial applications, such as the CZ wafer manufacture. Simulation results show that it will compare well with the human performance. 1. Introduction During the widely used CZ semiconductor production process, highly irregular shaped polycrystalline silicon nuggets are packed into a large quartz crucible, see Figure 1 [7]. Each highly irregularly shaped nugget is unique, with weights ranging ...

Abstract—We define a utility-based framework for joint channel assignment and topology control in multi-rate multi-radio wireless mesh networks, and present a greedy algorithm for solving the corresponding optimization problem. Key features of the proposed approach are the support for different target objectives, which are expressed as utility functions of the MAC layer throughput, and the efficient utilization of wired network gateways, while guaranteeing that for every mesh node there exists a path to a gateway. Investigations show the influence of different target objectives on the channel assignment and network topology, demonstrate the proposed approach’s load balancing properties in mesh networks containing multiple gateways, show the effect of 802.11a adjacent channel interference on the channel assignment, and compare the performance of the proposed procedure with a rate-based channel assignment scheme. I.