V-SVR+: Support Vector Regression with Variational Privileged Information
- Publisher:
- Institute of Electrical and Electronics Engineers (IEEE)
- Publication Type:
- Journal Article
- Citation:
- IEEE Transactions on Multimedia, 2021, PP, (99), pp. 1-1
- Issue Date:
- 2021-01-01
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Many regression tasks encounter an asymmetric distribution of information between training and testing phases where the additional information available in training, the so-called privileged information, is often inaccessible in testing. In practice, the privileged information (PI) in training data might be expressed in different formats, such as continuous, ordinal, or binary values. However, most of the existing learning using privileged information (LUPI) paradigms mainly deal with the continuous form of PI, resulting in their incapability to handle variational PI, which motivates this research. Therefore, in this paper, we propose a unified framework to tackle the aforementioned three forms of privileged information systematically. The proposed method (called V-SVR+) integrates the continuous, ordinal, and binary PI into the learning process of support vector regression (SVR) via the proposed three losses. Specifically, for continuous privileged information, we define a linear correcting (slack) function in the privileged information space to estimate the slack variables in the standard SVR method using privileged information. For the ordinal relations of privileged information, we first rank the privileged information. Then, we regard this ordinal privileged information as auxiliary information applied in the learning process of the SVR model. For the binary or Boolean form of privileged information, we infer a probabilistic dependency between the privileged information and labels from the summarized privileged information knowledge. Then, we transfer the privileged information knowledge to constraints and form a constrained optimization problem. We evaluate the proposed method on three applications: music emotion recognition from songs with the help of implicit information about music elements judged by composers; multiple object recognition from images with the help of implicit information about the object's importance conveyed by the list of manually annotated image tags; and photo aesthetic assessment enhanced by high-level aesthetic attributes hidden in photos. The experiment results demonstrate that the proposed methods are superior to the classical learning paradigm when solving practical problems.
Please use this identifier to cite or link to this item: