Testing using Privileged Information by Adapting Features with Statistical Dependence
This addresses scenarios where training data or labels are unavailable or expensive, offering a method to enhance predictions incrementally without access to the original model.
The paper tackles the problem of improving an imperfect predictor at test time using additional features without retraining, by estimating and strengthening statistical dependence via manifold denoising, resulting in demonstrated improvement in real-world visual attribute ranking.
Given an imperfect predictor, we exploit additional features at test time to improve the predictions made, without retraining and without knowledge of the prediction function. This scenario arises if training labels or data are proprietary, restricted, or no longer available, or if training itself is prohibitively expensive. We assume that the additional features are useful if they exhibit strong statistical dependence to the underlying perfect predictor. Then, we empirically estimate and strengthen the statistical dependence between the initial noisy predictor and the additional features via manifold denoising. As an example, we show that this approach leads to improvement in real-world visual attribute ranking. Project webpage: http://www.jamestompkin.com/tupi