Filters








931 Hits in 2.9 sec

Cross-Validation and Minimum Generation Error based Decision Tree Pruning for HMM-based Speech Synthesis

Heng Lu, Zhen-Hua Ling, Li-Rong Dai, Ren-Hua Wang
2010 International Journal of Computational Linguistics and Chinese Language Processing  
Therefore, a minimum cross generation error (MCGE) based decision tree pruning method for HMM-based speech synthesis is proposed in this paper.  ...  This paper presents a decision tree pruning method for the model clustering of HMM-based parametric speech synthesis by cross-validation (CV) under the minimum generation error (MGE) criterion.  ...  Experiments on Decision Tree Pruning Objective Evaluation Using the threshold parameter pair {0.01,15} of Sys-D, the initial decision tree 0 TR was built by conducting MDL-based HMM clustering using  ... 
dblp:journals/ijclclp/0002LDW10 fatcat:zkrkpkgsxvg4njj763zhxm2kii

Building simple models: A case study with decision trees [chapter]

David Jensen, Tim Oates, Paul R. Cohen
1997 Lecture Notes in Computer Science  
Many approaches to decision tree induction fail this challenge.  ...  Adjusting for these effects produces trees with little or no excess structure.  ...  and costcomplexity pruning.  ... 
doi:10.1007/bfb0052842 fatcat:imiblmxrfzdu7b3mcoszia6kdy

Fast model adaptation and complexity selection for nonnative English speakers

Xiaodong He, Yunxin Zhao
2002 IEEE International Conference on Acoustics Speech and Signal Processing  
In MDL/PL, MDL is performed on nodes with sufficient adaptation data, and pseudolikelihood based state tying is performed on nodes with insufficient adaptation data.  ...  A novel technique of combining MDL with pseudo likelihood-based state-tying is proposed to enable model complexity selection from using as little as three adaptation speech sentences.  ...  In MDL-based model complexity selection, detailed phonetic decision trees are first built to organize allophone states hierarchically, and MDL is applied to perform tree pruning and the optimal tree cuts  ... 
doi:10.1109/icassp.2002.5743783 dblp:conf/icassp/HeZ02 fatcat:bjyehdabvbdfjmnbbt7rekntxy

A Comparison of Approaches for Learning Probability Trees [chapter]

Daan Fierens, Jan Ramon, Hendrik Blockeel, Maurice Bruynooghe
2005 Lecture Notes in Computer Science  
Probability trees (or Probability Estimation Trees, PET's) are decision trees with probability distributions in the leaves.  ...  In this paper we experimentally compare the main approaches using the relational decision tree learner Tilde (both on non-relational and on relational datasets).  ...  As an alternative we can use T hr = 0 and apply post-pruning based on MDL-reasoning [5] . We refer to this approach as MDLp. MDLp builds trees at least as large as those for MDLs.  ... 
doi:10.1007/11564096_54 fatcat:wqxgvhzp2zdntkkux2onnmzvbm

Inducing decision trees with an ant colony optimization algorithm

Fernando E.B. Otero, Alex A. Freitas, Colin G. Johnson
2012 Applied Soft Computing  
tree induction, and the accuracy of the ACO-based cACDT decision tree algorithm.  ...  In this paper we propose a novel ACO algorithm to induce decision trees, combining commonly used strategies from both traditional decision tree induction algorithms and ACO.  ...  procedure and the two-step tree pruning, denoted by Ant-Tree-Miner MDL ; 4. using a MDL-based discretisation procedure and only C4.5's error-based pruning procedure, denoted by Ant-Tree-Miner -P MDL .  ... 
doi:10.1016/j.asoc.2012.05.028 fatcat:3wyxkn2kgrb5zaiuxait2icvxi

A comparison of pruning criteria for probability trees

Daan Fierens, Jan Ramon, Hendrik Blockeel, Maurice Bruynooghe
2009 Machine Learning  
Probability trees are decision trees that predict class probabilities rather than the most likely class.  ...  The main conclusion is that overall a pruning criterion based on randomization tests performs best because it is most robust to extreme data characteristics (such as class skew or a high number of classes  ...  Error Based Pruning (EBP) 3.5.1 Main Idea Error Based Pruning (EBP) was developed for classification trees.  ... 
doi:10.1007/s10994-009-5147-1 fatcat:li7yqslldzai3a2sgbsa622uve

SLIQ: A fast scalable classifier for data mining [chapter]

Manish Mehta, Rakesh Agrawal, Jorma Rissanen
1996 Lecture Notes in Computer Science  
SLIQ also uses a new tree-pruning algorithm that is inexpensive, and results in compact and accurate trees.  ...  SLIQ is a decision tree classi er that can handle both numeric and categorical attributes. It uses a novel pre-sorting technique in the tree-growth phase.  ...  In MDL Pruning Section 4.4 presented the partial and hybrid MDL-based pruning algorithms that can remove a subset of the children at any decision tree node.  ... 
doi:10.1007/bfb0014141 fatcat:dk2fhorrajb4bfej3kamt4empi

Oblique Decision Trees using embedded Support Vector Machines in classifier ensembles

Vlado Menkovski, Ioannis T. Christou, Sofoklis Efremidis
2008 2008 7th IEEE International Conference on Cybernetic Intelligent Systems  
We present a new base classifier that utilizes Oblique Decision Tree technology based on Support Vector Machines for the construction of oblique (non-axis parallel) tests on the nodes of the decision tree  ...  To overcome this problem and achieve better generalization a mechanism for pruning the decision tree is implemented.  ... 
doi:10.1109/ukricis.2008.4798937 fatcat:zorjt726yvhctlpt6c7fsp6pxe

Adaptive Channel Equalization Using Classification Trees

Taneli Haverinen, Arto Kantsila, Mikko Lehtokangas, Jukka Saarinen
2015 Zenodo  
We have used MDL-SLIQ algorithm [7] , which is based on the idea of finding the minimum description length (MDL) of the objects' classes in the training data.  ...  Black solid lines represent trees prior to pruning, gray lines represent pruned trees. BER of neural network equalizer is plotted with dashed line.  ... 
doi:10.5281/zenodo.37182 fatcat:dnkvvveiknhnpflkdmkux6rrxm

Maximum a posteriori pruning on decision trees and its application to bootstrap BUMPing

Jinseog Kim, Yongdai Kim
2006 Computational Statistics & Data Analysis  
On the other hand, the pruning algorithms based on posterior calculations such as BIC (MDL) and MEP are faster, but they sometimes produce too big or small trees to yield poor generalization errors.  ...  pruning.  ...  Review of Pruning Methods In this section, we briefly review the three pruning methods based on CCP, BIC/MDL and MEP, respectively.  ... 
doi:10.1016/j.csda.2004.09.010 fatcat:ess3bjwotrcr7cpjpbdfqdokea

Supervised ranking in open-domain text summarization

Tadashi Nomoto, Yuji Matsumoto
2001 Proceedings of the 40th Annual Meeting on Association for Computational Linguistics - ACL '02  
We build probabilistic decision trees of different flavors and integrate each of them with the clustering framework.  ...  In particular, we explore the use of probabilistic decision tree within the clustering framework to account for the variation as well as regularity in human created summaries.  ...  MDL-DT MDL-DT stands for a decision tree with MDL based pruning.  ... 
doi:10.3115/1073083.1073161 dblp:conf/acl/NomotoM02 fatcat:l5c7cqqg4ndqfj5kbfzz5u6t4a

Advances in decision tree construction

Johannes Gehrke, Wei-Yin Loh
2001 Tutorial notes of the seventh ACM SIGKDD international conference on Knowledge discovery and data mining - KDD '01  
covered) MDL pruning Pruning by randomization testing T3-16 and Loh KDD 2001 Tutorial: Advances in Decision Trees Tutorial: Advances in Decision Trees Pruning Using the MDL Principle (Mehta, Rissanen  ...  Resulting tree is the final tree output by the pruning algorithm.Gehrke and Loh KDD 2001 Tutorial: Advances in Decision Trees Performance Improvements: PUBLIC (Shim and Rastogi, VLDB 1998) MDL bottom-up  ... 
doi:10.1145/502786.502790 fatcat:vvgrqggvdbhk5kmgfx5f4ockim

Modeling (in)variability of human judgments for text summarization

Tadashi Nomoto, Yuji Matsumoto
2002 Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval - SIGIR '02  
In particular, we explore the use of probabilistic decision tree within the clustering framework to account for the variation as well as regularity in human created summaries.  ...  MDL-DT is like C4.5 except that it makes use of an optimized pruning based on Minimum Description Length Principle (MDL) in place of reduced error pruning used in C4.5.  ...  To see this, we consider three decision tree algorithms of different flavors: C4.5, MDL-DT, and SSDT.  ... 
doi:10.1145/564376.564467 dblp:conf/sigir/NomotoM02 fatcat:fdjeloaanjgyrobgevkpzdzjza

A parallel two-pass MDL context tree algorithm for universal source coding

Nikhil Krishnan, Dror Baron, Mehmet Kivanc Mihcak
2014 2014 IEEE International Symposium on Information Theory  
Instead, our approach is to first estimate the minimum description length (MDL) source underlying the entire input, and then encode each of the B blocks in parallel based on the MDL source.  ...  Its redundancy is approximately B log(N/B) bits above Rissanen's lower bound on universal coding performance, with respect to any tree source whose maximal depth is at most log(N/B).  ...  Details of the pruning decision appear in Section III-C3.  ... 
doi:10.1109/isit.2014.6875156 dblp:conf/isit/KrishnanBM14 fatcat:nte6szi6nbbn7c7aw4dpp7jpei

The Method Research of Knowledge Discovery in Reservoir Management

Yao Jun, Zhang Yun
2006 2006 Semantics, Knowledge and Grid, Second International Conference on  
So we should make use of some particular data mining methods to discover knowledge, which should be based on some particular fields.  ...  Pruning is done by replacing a whole sub-tree by a leaf node. The pruning strategy used is based on the principle of MDL.  ...  The research of knowledge discovery that based on decision tree Decision tree is a valid method that can discover concept description space.  ... 
doi:10.1109/skg.2006.103 dblp:conf/skg/YaoZ06 fatcat:4jkk4qphnrhezddynp4vxeamxq
« Previous Showing results 1 — 15 out of 931 results