We propose a new parallel learning algorithm of latent local support vector machines (SVM), called latent-lSVM for effectively classifying very high-dimensional and large-scale multi-class datasets. The common framework of texts/images classification tasks using the Bag-Of-(visual)-Words model for the data representation leads to hard classification problem with thousands of dimensions and hundreds of classes. Our latent-lSVM algorithm performs these complex tasks into two main steps. The first one is to use latent Dirichlet allocation for assigning the datapoint (text/image) to some topics (clusters) with the corresponding probabilities. This aims at reducing the number of classes and the number of datapoints in the cluster compared to the full dataset, followed by the second one: to learn in a parallel way nonlinear SVM models to classify data clusters locally. The numerical test results on nine real datasets show that the latent-lSVM algorithm achieves very high accuracy compared to state-of-the-art algorithms. An example of its effectiveness is given with an accuracy of 70.14% obtained in the classification of Book dataset having 100 000 individuals in 89 821 dimensional input space and 661 classes in 11.2 minutes using a PC Intel(R) Core i7-4790 CPU, 3.6 GHz, 4 cores.
KEYWORDSLatent Dirichlet allocation (LDA), high-dimensional and large-scale multi-class data classification, parallel learning on multi-core computers, support vector machines (SVMs)
INTRODUCTIONThere are more and more multimedia data stored electronically, with increasing number of internet users and mobile internet access sharing videos, songs, or photos. There are more than 1 billion daily active users-nearly one-third of all people on the Internet (around 46% of the world population)-on Youtube and Facebook (Amazon and Yahoo! have even more), 600 000 hours (68 years) of videos are uploaded on Youtube every day, and 46 000 years are viewed at the same time. Almost all mobile phones can take photos: 2 trillions photos will be shared this year. There are 310 millions Twitter users and more than 600 millions Weibo users (the "Chinese Twitter';' Asia is the first Internet region with more than 50% of Internet users in 2016). The number of data is always increasing and their sizes too: 4K or 3D videos, sound in Dolby 5.1, higher and higher photo resolution, text messages replaced by voice messages. This leads to very huge amount of data; there is a need for high performance classification algorithms in order to help us find what we are looking for. We present a new fast and accurate parallel local support vector machine (SVM) algorithm for the classification of very large scale and high-dimensional multi-class datasets. The experimental results are performed on two different kinds of datasets: image and text classification.The classification of texts/images is one of the important research topics in text mining, computer vision, and machine learning. The purpose is to ask a computer to assign the predefined class label to the text/image. The popu...