Adaptive Elastic Input Field for Recognition Improvement
نویسنده
چکیده
For machines to perform classification tasks, such as speech and character recognition, appropriately handling deformed patterns is a key to achieving high performance. The authors presents a new type of classification system, an Adaptive Input Field Neural Network (AIFNN), which includes a simple pre-trained neural network and an elastic input field attached to an input layer. By using an iterative method, AIFNN can determine an optimal affine translation for an elastic input field to compensate for the original deformations. The convergence of the AIFNN algorithm is shown. AIFNN is applied for handwritten numerals recognition. Consequently, 10.83% of originally misclassified patterns are correctly categorized and total performance is improved, without modifying the neural network.
منابع مشابه
روشی نو برای حذف پژواک آکوستیکی استریو با استفاده از ساختار مبتنی بر الگوریتم وفقی با ورودی برش یافته
Stereophonic acoustic echo cancellation is one of the expanding areas in the field of the speech/multimedia communication systems. In a conventional stereophonic acoustic echo canceller, working in a stereo communication system, the existence of the strong cross-correlation between the input signals to the two channels has the problem of the low convergence speed of the weights of the adaptive ...
متن کاملImproving adaptive resolution of analog to digital converters using least squares mean method
This paper presents an adaptive digital resolution improvement method for extrapolating and recursive analog-to-digital converters (ADCs). The presented adaptively enhanced ADC (AE-ADC) digitally estimates the digital equivalent of the input signal by utilizing an adaptive digital filter (ADF). The least mean squares (LMS) algorithm also determines the coefficients of the ADF block. In this sch...
متن کاملTurbo Expander System Behavior Improvement Using an Adaptive Fuzzy PID Controller
Turbo-expanders are used in industries for cooling, liquefaction and also power generation. An important part of these turbines is the variable angle nozzle causing a nonlinear behavior that is not well recognized among the prime movers of the dispersed generators. In this paper, at first, the turbo expander system is evaluated in details and its nonlinear behavior is investigated. Then, the sy...
متن کاملHand Written Character Recognition Using Dynamic Reshaping
In this project, we propose a new approach to hand written character recognition, called dynamic reshaping, which avoids the respective drawbacks of classical neural network approaches. Contrary to the neural network based recognition, our scheme does not need any training set, instead, only very limited character templates are needed. Dynamic reshaping recognition is based on the dynamic trans...
متن کاملRobust Adaptive Attitude Stabilization of a Fighter Aircraft in the Presence of Input Constraints
The problem of attitude stabilization of a fighter aircraft is investigated in this paper. The practical aspects of a real physical system like existence of external disturbance with unknown upper bound and actuator saturation are considered in the process of controller design of this aircraft. In order to design a robust autopilot in the presence of the actuator saturation, the Composite Nonli...
متن کامل