Ultraconservative Online Algorithms for Multiclass Problems

By Koby Crammer, Yoram Singer. Journal of machine learning research, 2003

742 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/28b9bacde6499f8cc6f7e70feee4232107211e39

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Learning with Compositional Semantics as Structural Inference for Subsentential Sentiment Analysis (2008, methodology): “In our experiment, we use an online SVM algorithm called MIRA (Margin Infused Relaxed Algorithm) (Crammer and Singer, 2003) 7 for training.”
An Introduction to Conditional Random Fields (2010, methodology): “The online perceptron update can also be made margin-aware, yielding the MIRA algorithm [23], which may perform better than the perceptron update.”
Label ranking by learning pairwise preferences (2008, methodology): “For example, Crammer and Singer (2003b) consider a variety of on-line learning algorithms for the problem of ranking possible labels in a multi-label text categorization task.”
Online Multitask Learning (2006, methodology): “An analysis of the L∞ norm of prediction errors is implicit in some past work of Crammer and Singer [11, 12]; the present paper extends this work to a broader framework, and tightens the analysis.”
Social big data: Recent achievements and new challenges (2015, methodology): “A new generation of online [132,133] and streaming [134,135] alorithms is currently being developed in order to manage social big ata challenges, and these algorithms require high scalability in both emory consumption [136] and time computation.”
Online Learning of Multiple Tasks with a Shared Loss (2007, methodology): “An analysis of the L∞ norm of prediction errors is implicit in some past work of Crammer and Singer (2001, 2003).”
Cutting-plane training of structural SVMs (2009, methodology): “…prediction, the PEGASOS algorithm (Shalev-Shwartz et al. 2007) has shown promising performance for binary classiﬁcation SVMs. Related to such online methods is also the MIRA algorithm (Crammer and Singer 2003), which has been used for training structured predictors (e.g. McDonald et al. 2005).”
Efficient bandit algorithms for online multiclass prediction (2008, methodology): “Our starting point is a simple adaptation of the Perceptron algorithm [Rosenblatt, 1958] for multiclass prediction in the full information case (this adaptation is called Kesler s construction in [Duda and Hart, 1973, Crammer and Singer, 2003]).”
Sentence Compression by Deletion with LSTMs (2015, methodology): “Compression ratio: The three versions of our system ( LSTM *) and the baseline ( MIRA ) have comparable compression ratios (CR) which are deﬁned as the length of the compression in characters divided over the sentence length.”
A comparison of methods for multiclass support vector machines (2002, methodology): “In [7], an O(k log k) algorithm was proposed for (18) while in [8], a simple iterative approach was used.”
Online Machine Learning in Big Data Streams (2018, methodology): “ation procedures were proposed prior to PA: Kivinen and Warmuth [135] give an overview of numerous earlier additive and multiplicative online algorithms, all of which are solvable by gradient descent [160, 100, 62, 134]. Based on the experiments in [62, 60], the Passive Aggressive algorithm outperfo”
Structured Models for Fine-to-Coarse Sentiment Analysis (2007, methodology): “The weights, w, are set using the MIRA learning algorithm, which is an inference based online largemargin learning technique (Crammer and Singer, 2003; McDonald et al., 2005).”
Discourse Constraints for Document Compression (2010, methodology): “The model parameters are estimated using the Margin Infused Relaxed Algorithm (MIRA; Crammer and Singer 2003), a discriminative large-margin online learning technique.”
A survey on opinion mining and sentiment analysis: Tasks, approaches and applications (2015, unknown): “algorithm was employed for classification purpose [10, 20].”
Fast Kernel Classifiers with Online and Active Learning (2005, methodology): “Other algorithms (Frieß et al., 1998; Gentile, 2001; Li and Long, 2002; Crammer and Singer, 2003) explicitely construct larger margins.”
A dual coordinate descent method for large-scale linear SVM (2008, methodology): “The algorithm for multi-class problems in (Crammer & Singer, 2003) is also similar to our online setting.”