We present a novel multitask learning framework called multitask generalized eigenvalue program (MTGEP), which jointly solves multiple related generalized eigenvalue problems (GEPs). This framework is quite general and can be applied to many eigenvalue problems in machine learning and pattern recognition, ranging from supervised learning to unsupervised learning, such as principal component analysis (PCA), Fisher discriminant analysis (FDA), common spatial pattern (CSP), and so on. The core assumption of our approach is that the leading eigenvectors of related GEPs lie in some subspace that can be approximated by a sparse linear combination of basis vectors. As a result, these GEPs can be jointly solved by a sparse coding approach. Empirical evaluation with both synthetic and benchmark real world datasets validates the efficacy and efficiency of the proposed techniques, especially for grouped multitask GEPs.

Semi-supervised learning (SSL) is an important research problem in machine learning. While it is usually expected that the use of unlabeled data can improve performance, in many cases SSL is outperformed by supervised learning using only labeled data. To this end, the construction of a performance-safe SSL method has become a key issue of SSL study. To alleviate this problem, we propose in this paper the UMVP (safe semi-sUpervised learning for MultiVariate Performance measure) method, because of the need of various performance measures in practical tasks. The proposed method integrates multiple semi-supervised learners, and maximizes the worst-case performance gain to derive the final prediction. The overall problem is formulated as a maximin optimization. In oder to solve the resultant difficult maximin optimization, this paper shows that when the performance measure is the Top- k Precision, F β score or AUC, a minimax convex relaxation of the maximin optimization can be solved efficiently. Experimental results show that the proposed method can effectively improve the safeness of SSL under multiple multivariate performance measures.

We study the robustness of active learning (AL) algorithms against prior misspecification: whether an algorithm achieves similar performance using a perturbed prior as compared to using the true prior. In both the average and worst cases of the maximum coverage setting, we prove that all alpha-approximate algorithms are robust (i.e., near alpha-approximate) if the utility is Lipschitz continuous in the prior. We further show that robustness may not be achieved if the utility is non-Lipschitz. This suggests we should use a Lipschitz utility for AL if robustness is required. For the minimum cost setting, we can also obtain a robustness result for approximate AL algorithms. Our results imply that many commonly used AL algorithms are robust against perturbed priors. We then propose the use of a mixture prior to alleviate the problem of prior misspecification. We analyze the robustness of the uniform mixture prior and show experimentally that it performs reasonably well in practice.

A downside of K-Nearest Neighbors is that you need to hang on to your entire training dataset. The Learning Vector Quantization algorithm (or LVQ for short) is an artificial neural network algorithm that lets you choose how many training instances to hang onto and learns exactly what those instances should look like. In this post you will discover the Learning Vector Quantization algorithm. This post was written for developers and assumes no background in statistics or mathematics. The post focuses on how the algorithm works and how to use it for predictive modeling problems.

Illinois Structured Learning Package (Illinois-SL) is a general purpose JAVA library for performing structured learning. It houses learning algorithms like averaged Structured Perceptron and Structured SVM with L2-Loss, and provides a minimal interface for your structured learning needs. The training algorithm employed for training SSVM is dual coordinate descent(DCD), which has been proven to have very good convergence properties. Illinois-SL comes with an efficient implementation of DCD with support for multi-threading. Illinois-SL provides a simple and neat framework for developing applications using structured prediction models.

Relation extraction, which learns semantic relations of concept pairs from text, is an approach for mining commonsense knowledge. This paper investigates an approach for relation extraction, which helps expand a commonsense knowledge base with little labor work. We proposed a framework that learns new pairs from Chinese corpora by adopting concept pairs in Chinese commonsense knowledge base as seeds. Multiple instance learning is utilized as the learning algorithm for predicting relation for unseen pairs. The performance of our system could be improved by learning multiple iterations. The results in each iteration are manually evaluated and processed to next iteration as seeds. Our experiments extracted new pairs for relations “AtLocation”, “CapableOf”, and “HasProperty”. This study showed that new pairs could be extracted from text without huge humans work.

On Thursday, Aaron's Inc., a Maryland-based appliance and electronics company, set a new Guinness World Record for the "largest human mattress dominoes" chain with 1,200 participants taking a total of 13 minutes and 38 seconds to complete the larger than life feat. Event organizers used two exhibit halls covering 70,000 square feet to set up 34 rows of mattresses. The first mattress was pushed over by Aaron's CEO John Robinson. "Breaking a Guinness World Records title has been a great team building event for the associates we have attending our National Managers meeting," said Robinson at the event. The event not only broke a world record but supported a great cause.

Our matching problem setting is similar to the transductive setting for classification, from Vapnik [9], where there is a set of training examples with known inputs and class labels and a set of working examples with known inputs and unknown class labels, and the goal is to use the available training and working data to develop a classifier that classifies the working examples with a low error rate. For results on validation of network classifiers (rather than reconciliation algorithms) in transductive settings, refer to [10] and [11]. For theory and insight on why collective classification succeeds in general settings and validation methods for it, refer to [12]. For network reconciliation, we assume that we know some network data, consisting of some node data and the links, for both networks involved in the matching, and our goal is to use that network data to match nodes as accurately as possible between the networks. This paper presents a technique to compute probably approximately correct (PAC) bounds on the precision and recall of matching algorithms.

Here, we discuss some methods for carrying out classification when only positive examples are available. The latter half of our discussion borrows heavily from W.S. Lee and B. Liu, Proc. Follow @efavdb Follow us on twitter for new submission alerts! Logistic regression is a commonly used tool for estimating the level sets of a Boolean function y on a set of feature vectors \textbf{F}: In a sense, you can think of it as a method for playing the game "Battleship" on whatever data set you're interested in. Consider now a situation where all training examples given are positive -- i.e., no negative examples are available.

What is supervised machine learning and how does it relate to unsupervised machine learning? In this post you will discover supervised learning, unsupervised learning and semis-supervised learning. Supervised and Unsupervised Machine Learning Algorithms Photo by US Department of Education, some rights reserved. The majority of practical machine learning uses supervised learning. Supervised learning is where you have input variables (x) and an output variable (Y) and you use an algorithm to learn the mapping function from the input to the output.