Abstract :
Adaptive feature learning is an effective method to explore the mechanism of information pattern recognition in information flow. This paper integrates the progresses of expert learning and artificial intelligence to propose a few new learning algorithms for pattern recognition in information flow. For solving high order matrix computing problem, this paper proposes an orthogonal transformation algorithm. For solving frequency modulation (FM) pattern recognition problem, this paper proposes a differential algorithm. For solving unknown pattern recognition in large scale information flow problem, this paper proposes inverse convolution algorithm and probability spectrum algorithm. These feature learning algorithms can extract and recognize pattern fast, efficiently and explicitly, even patterns are complex, confused and incomplete.
Keywords :
learning (artificial intelligence); matrix algebra; pattern recognition; adaptive feature learning; artificial intelligence; expert learning; frequency modulation pattern recognition problem; information flow; information pattern recognition; inverse convolution algorithm; orthogonal transformation algorithm; probability spectrum algorithm; Artificial intelligence; Convolution; Eigenvalues and eigenfunctions; Frequency modulation; IEEE members; Large-scale systems; Learning; Noise reduction; Pattern recognition; Telecommunication computing; Eigen value; Feature Extraction; High Power Matrix Computing; Orthogonal Transformation; inverse convolution; probability spectrum;
