| 摘要 |
The mapping into a space of reduced dimensionality may be accomplished by a linear mapping algorithm, which may be the principal component analysis (PCA) algorithm, or a variant thereof. The standardising transformation may be performed using a multiple linear regression algorithm. Alternatively, the standardising transformation may be performed using a multiple non-linear regression algorithm, a restricted mathematical transformation or a neural network. The input pattern may comprise the response of an array of gas sensors or quantities related thereto. The response may be normalised before being inputted to the neural network, or mapped to a space of reduced dimensionality. |
