Title:
【発明の名称】神経回路網を用いた波形評価装置
Document Type and Number:
Japanese Patent JP3378996
Kind Code:
B2
Abstract:
PURPOSE:To easily evaluate a waveform without using any complicated program by learning the neuron superposition rate of a neural network module through a model teacher of an ideal waveform signal and a bad-modal teacher of an undesirable waveform signal. CONSTITUTION:The waveform data to be collected are inputted to a waveform statistic collecting part 21 from a measured subject 10, and the part 21 performs the waveform statistic processing to form an ideal waveform. Then, the part 21 decides the waveform to be collected and collects it. Then, a neural network module to be learnt is selected. If a phase error is decided, for example, a waveform of a typical phase is selected at a teacher waveform deciding part 22 as a model teacher together with a typical example of an undesirable phase selected as a bad-model teacher, respectively. A waveform identifying part 23 calculates the evaluating function of an output when a model teacher waveform is inputted. If the evaluating function is out of a prescribed range, the neuron superposition rate is corrected based on the neuron superposition rate correction value. This routine is repeated until the neuron superposition rate with which the evaluating function value is set within the prescribed range is decided.
Inventors:
Nobuo Takachi
Application Number:
JP15598492A
Publication Date:
February 17, 2003
Filing Date:
May 23, 1992
Export Citation:
Assignee:
Topcon Co., Ltd.
International Classes:
G06F15/18; G06G7/60; G06N3/04; G06N99/00; (IPC1-7): G06G7/60; G06N1/00
Domestic Patent References:
| JP3118605A | ||||
| JP447378A |
Attorney, Agent or Firm:
Yuichi Izumi