Controling the Magniication Factor of Self-organizing Feature Maps
نویسندگان
چکیده
The magniication exponents occuring in adaptive map formation algorithms like Kohonen's self-organizing feature map deviate for the information theoretically optimal value = 1 as well as from the values which optimize, e.g., the mean square distortion error (= 1=3 for one-dimensional maps). At the same time, models for categorical perception such as the \perceptual magnet" eeect which are based on topographic maps require negative mag-niication exponents < 0. We present an extension of the self-organizing feature map algorithm which utilizes adaptive local learning step sizes to actually control the magniication properties of the map. By change of a single parameter, maps with optimal information transfer, with various minimal reconstruction errors, or with an inverted magniication can be generated. Analytic results on this new algorithm are complemented by numerical simulations .
منابع مشابه
Steel Consumption Forecasting Using Nonlinear Pattern Recognition Model Based on Self-Organizing Maps
Steel consumption is a critical factor affecting pricing decisions and a key element to achieve sustainable industrial development. Forecasting future trends of steel consumption based on analysis of nonlinear patterns using artificial intelligence (AI) techniques is the main purpose of this paper. Because there are several features affecting target variable which make the analysis of relations...
متن کاملTowards an Information Density Measure for Neural Feature Maps
Many neural models have been suggested for the development of feature maps in cortical areas. Undoubtedly the most popular model is the Kohonen self-organizing map (SOM). Once the map has been learned, this network uses a competitive winner-take-all (WTA) approach to choose a singlèbest' output neuron on a (typically) 2D grid for each presented input pattern. Cortical maps in biological organis...
متن کاملMagni cation in Neural Information Channel Maps
Many cortical maps are found in biological organisms. In these maps, related units tend to be collected close to each other, and important inputs typically have a magniied representation on the map. We consider an interpretation of these structures as maps of parallel information-bearing channels, rather than winner-take-all maps such as the Kohonen self-organizing map (SOM). By considering the...
متن کاملمنطقه بندی حوزه های آبخیز با به کارگیری نوعی از شبکه های عصبی مصنوعی به منظور تحلیل فراوانی منطقه ای سیلاب
Self-Organizing Feature Maps (SOFM) are a variety of artificial neural networks that their applications in the areas of pattern recognition and data clustering makes them noticeable tools to perform regional flood frequency analysis (RFFA). In this study, ability of Self-Organizing Feature Maps for regionalization of Sefidrood watershed in order to perform regional flood frequency analysis usin...
متن کامل