| |
| Sparse discriminant analysis (SDA) imposes $l$-1 regularization to encourage sparse coefficients in linear discriminant transform. This approach has found a broad range of machine learning tasks, due to its capability of identifying the most promising features so that the feature dimensionality can be significantly reduced, leading to most robust and generalizable models. This paper reviews the development of SDA from linear discriminative analysis (LDA), and presents its application to the driving distraction detection task. |
|
| Keywords:Machine learning, signal processing, distraction detection, sparse discriminative analysis |
|
