Neural Classifiers for Learning Higher-Order Correlations
نویسنده
چکیده
Studies by various authors suggest that higher-order networks can be more powerful and are biologically more plausible with respect to the more traditional multilayer networks. These architectures make explicit use of nonlinear interactions between input variables in the form of higher-order units or product units. If it is known a priori that the problem to be implemented possesses a given set of invariances like in the translation, rotation, and scale invariant pattern recognition problems, those invariances can be encoded, thus eliminating all higher-order terms which are incompatible with the invariances. In general, however, it is a serious setback that the complexity of learning increases exponentially with the size of inputs. This paper reviews higher-order networks and introduces an implicit representation in which learning complexity is mainly decided by the number of higher-order terms to be learned and increases only linearly with the input size.
منابع مشابه
A Higher Order Online Lyapunov-Based Emotional Learning for Rough-Neural Identifiers
o enhance the performances of rough-neural networks (R-NNs) in the system identification, on the base of emotional learning, a new stable learning algorithm is developed for them. This algorithm facilitates the error convergence by increasing the memory depth of R-NNs. To this end, an emotional signal as a linear combination of identification error and its differences is used to achie...
متن کاملLearning Document Image Features With SqueezeNet Convolutional Neural Network
The classification of various document images is considered an important step towards building a modern digital library or office automation system. Convolutional Neural Network (CNN) classifiers trained with backpropagation are considered to be the current state of the art model for this task. However, there are two major drawbacks for these classifiers: the huge computational power demand for...
متن کاملDesigning Kernel Scheme for Classifiers Fusion
In this paper, we propose a special fusion method for combining ensembles of base classifiers utilizing new neural networks in order to improve overall efficiency of classification. While ensembles are designed such that each classifier is trained independently while the decision fusion is performed as a final procedure, in this method, we would be interested in making the fusion process more a...
متن کاملA model with an intrinsic property of learning higher order correlations
A neural network model that can learn higher order correlations within the input data without suffering from the combinatorial explosion problem is introduced. The number of parameters scales as M x N, where M is the number such that no higher order network with less than M higher order terms can implement the same input data set and N is the dimensionality of the input vectors. In order to hav...
متن کاملNeural Network Classifiers and Gene Selection Methods for Microarray Data on Human Lung Adenocarcinoma
In an attempt to account for correlations in gene expression data, we considered neural network classifiers with random weights selected from a normal distribution. The optimal parameters of the distribution were determined using Bayesian methods. The performance of such a Bayesian neural network was compared with that of a standard feed-forward hierarchical neural network. The performance of t...
متن کامل