Efficient sampling of training set in large and noisy multimedia data

Abstract

As the amount of multimedia data is increasing day-by-day thanks to less expensive storage devices and increasing numbers of information sources, machine learning algorithms are faced with large-sized and noisy datasets. Fortunately, the use of a good sampling set for training influences the final results significantly. But using a simple random sample (SRS) may not obtain satisfactory results because such a sample may not adequately represent the large and noisy dataset due to its blind approach in selecting samples. The difficulty is particularly apparent for huge datasets where, due to memory constraints, only very small sample sizes are used. This is typically the case for multimedia applications, where data size is usually very large. In this article we propose a new and efficient method to sample of large and noisy multimedia data. The proposed method is based on a simple distance measure that compares the histograms of the sample set and the whole set in order to estimate the representativeness of the sample. The proposed method deals with noise in an elegant manner which SRS and other methods are not able to deal with. We experiment on image and audio datasets. Comparison with SRS and other methods shows that the proposed method is vastly superior in terms of sample representativeness, particularly for small sample sizes although time-wise it is comparable to SRS, the least expensive method in terms of time. ACM Reference Format:Wang, S., Dash, M., Chia, L.-T., and Xu, M. 2007. Efficient sampling of training set in large and noisy multimedia data.

Multimodal fusion for multimedia analysis: a survey

“…There are various solutions to deal with the noisy data. For instance, to employ some noise filter mechanisms to smooth the noisy data or to apply an appropriate sampling technique to differentiate the noisy data from the input data before the fusion takes place [137].…”

A Web-Based Multimedia Retrieval System with MCA-Based Filtering and Subspace-Based Learning Algorithms

“…In (Tseng et al, 2008), collaborative filtering is applied to the intelligent multimedia recommender based on the assumption that people can be grouped to share their valuable purchase information with each other to make the choices about the preferred items, such as movies and books. In recent years, several categories of data filtering methods have been investigated for multimedia retrieval such as algorithms using posterior probability (Angelova, Abu-Mostafa & Perona, 2005;Vezhnevets & Barinova, 2007), techniques based on distance or density measure (Chen et al, 2007;Wang, Dash, Chia, & Xu, 2007), and clustering based data filtering methods (Sarawagi, Deshpande & Kasliwal, 2009;Xiong, Gaurav, Steinbach & Vipin, 2006).…”

RANSAC-based training data selection for speaker state recognition

“…Their method is based on the Nearest Neighbor classifier. Wang et al [5], present a method to sample a large and noisy multimedia data. Their method is based on a simple distance measure that compares the histograms of the sample set and the whole set in order to assess the representativeness of the sample set.…”