One-Pass Boosting
نویسندگان
چکیده
This paper studies boosting algorithms that make a single pass over a set of base classifiers. We first analyze a one-pass algorithm in the setting of boosting with diverse base classifiers. Our guarantee is the same as the best proved for any boosting algorithm, but our one-pass algorithm is much faster than previous approaches. We next exhibit a random source of examples for which a “picky” variant of AdaBoost that skips poor base classifiers can outperform the standard AdaBoost algorithm, which uses every base classifier, by an exponential factor. Experiments with Reuters and synthetic data show that one-pass boosting can substantially improve on the accuracy of Naive Bayes, and that picky boosting can sometimes lead to a further improvement in accuracy.
منابع مشابه
Word Set Probability Boosting : The AB / TAB Algorithms
Based on the observation that the unpredictable nature of conversational speech makes it almost impossible to reliably model sequential word constraints, the notion of word set error criteria is proposed for improved recognition of spontaneous dialogues. The single pass Adaptive Boosting (AB) algorithm enables the language model weights to be tuned using the word set error criteria. In the two ...
متن کاملAn investigation into the efficacy of methods commonly employed by mix engineers to reduce frequency masking in the mixing of multitrack musical recordings
Studio engineers use a variety of techniques to reduce frequency masking between instruments when mixing multitrack musical recordings. This study evaluates the efficacy of three techniques, namely mirrored equalization, frequency spectrum sharing and stereo panning, against their variations to confirm the veracity of accepted practice. Mirrored equalisation involves boosting one instrument and...
متن کاملImproved Boosting Performance by Explicit Handling of Ambiguous Positive Examples
Visual classes naturally have ambiguous examples, that are different depending on feature and classifier and are hard to disambiguate from surrounding negatives without overfitting. Boosting in particular tends to overfit to such hard and ambiguous examples, due to its flexibility and typically aggressive loss functions. We propose a two-pass learning method for identifying ambiguous examples a...
متن کاملEvaluation of Data Mining Algorithms for Detection of Liver Disease
Background and Aim: The liver, as one of the largest internal organs in the body, is responsible for many vital functions including purifying and purifying blood, regulating the body's hormones, preserving glucose, and the body. Therefore, disruptions in the functioning of these problems will sometimes be irreparable. Early prediction of these diseases will help their early and effective treatm...
متن کاملDeep Discount Group Pass Programs: Innovative Transit Finance
For public transit operators in the U.S., neither fare increases nor fare reductions have been successful in boosting revenues. A different kind of strategy is needed, one that can produce more revenue for transit operators than it costs. This article argues that deep discount group pass (DDGP) programs can accomplish this goal. DDGP programs provide groups of people with unlimited-ride transit...
متن کامل