New Regularized Algorithms for Transductive Learning

By Partha P. Talukdar, Koby Crammer. 2009

241 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Revisiting Semi-Supervised Learning with Graph Embeddings (2016, methodology): “Transductive learning usually performs better than inductive learning because transductive learning can leverage the unlabeled test data when training the model (Joachims, 1999).”
Deeper Insights into Graph Convolutional Networks for Semi-Supervised Learning (2018, background): “The Laplacian smoothing (Taubin 1995) on each channel of the input features is deﬁned as: where 0 γ ≤ 1 is a parameter which controls the weighting between the features of the current vertex and the features of its neighbors.”
Structural Neighborhood Based Classification of Nodes in a Network (2016, background): “[21] proposed a modiﬁcation to this by deﬁning a well-behaved objective function which such a framework minimizes, thus guaranteeing convergence.”
Collective Classification in Network Data (2008, methodology): “An example of this is the modified adsorption algorithm [39].”
Large Scale Distributed Semi-Supervised Learning Using Streaming Approximation (2015, unknown): “The optimization criterion is inspired from [5] and similar to some existing approaches such as Adsorption [3] and MAD [29] but uses a slightly diﬀerent objective function, notably the matrices have diﬀerent constructions.”
Twitter Polarity Classification with Label Propagation over Lexical Links and the Follower Graph (2011, methodology): “We use the Junto Label Propagation Toolkit’s implementation of MAD in this paper.”
Semi-supervised User Geolocation via Graph Convolutional Networks (2018, background): “(2015a), which is based on location propagation using Modified Adsorption (Talukdar and Crammer, 2009), possibly because the label propagation in GCN is parametrised.”
Scaling Graph-based Semi Supervised Learning to Large Number of Labels Using Count-Min Sketch (2013, methodology): “In this section, we propose MAD-Sketch, a countmin sketch-based extension of MAD1 [17].”
A survey on semi-supervised learning (2019, methodology): “(2008) and Talukdar et al. (2008). The algorithm has been successfully applied to video recommendation, but is difficult to analyze theoretically, due to its many heuristic components.”
Adversarially Robust Generalization Just Requires More Unlabeled Data (2019, methodology): “For example, transductive SVM [16] uses prediction conﬁdence as a soft constraint, and graph-based SSL [3, 38] requires the model to have similar outputs at endpoints of an edge.”
Exploratory Learning (2013, methodology): “In the modified adsorption algorithm [25], one such graphbased label propagation method, each datapoint can be marked with one or more known labels, or a special dummy label meaning “none of the above”.”
A survey on classification techniques for opinion mining and sentiment analysis (2017, methodology): “2006), modified adsorption (Talukdar and Crammer 2009) and measure propagation (Subramanya and Bilmes 2011) are the most popular algorithms that have been used in this area.”
PIDGIN: ontology alignment using web text as interlingua (2013, methodology): “In PIDGIN, iterative matching is automatically obtained via the transitivity of inference in label propagation (MAD).”
Frame-Semantic Parsing (2014, background): “In our work, we are interested in multi-class generalizations of graph-propagation algorithms suitable for NLP applications, where each graph vertex can assume one or more out of many possible labels (Subramanya and Bilmes 2008, 2009; Talukdar and Crammer 2009).”