学术活动
Fuzzy Optimal Associate Memories for Modeling Chemical Profiles: Authentication of Foods and Nutrac
2013-11-18
来源：科技处 点击次数：主讲人：Peter de Boves Harrington
Ohio University Center for Intelligent Chemical Instrumentation
Department of Chemistry & Biochemistry
地 点：教四楼 406教室
主 办：化学系
Abstract：
Fuzzy Associative Optimal Memories (FOAMs) are content addressable memories that model the low-layer structure of the processing that occurs with the retina and optical nerves. Instead of treating a spectrum as a vector it is treated as an image. Basically a grid is placed over the spectrum and grid pixels are activated if a data point falls in the grid pixel. The grid is then unfolded and treated as a vector. A collection of these vectors are used to form an orthogonal basis that can be used as a one-class classifier. The advantage of this approach over other traditional classifiers such as Soft Independent Modeling for Class Analogies3 (SIMCA) is that the data are constrained with respect to the intensity axis using the gridded images of the FOAM method. By constraining the intensity axis, chance correlations are eliminated that occur when methods such as SIMCA or Principal Component Analysis are used.
