| Home | Download |  Service Center |  Jobs  | Add Favorites |

首頁 ＞ 人工智慧 ＞ 控制模擬 ＞ FlexTool for Matlab ＞ ENM
FlexTool(ENM): Evolutionary Neuro Modeling Tool
詳細功能:

  • 類神經 + 基因演算法對重覆的資料學習有不錯的展現。

  • ENM 是把結合類神經網路和遺傳發展演算。從數學科學到醫學的最佳化工具。FlexTool ( ENM ) 是為把遺傳發展演算法應用於多樣化領域的一個環境的套裝軟體。

  • FlexTool ( ENM ) 是在 MATLAB 的環境下操作的軟體。MATLAB 為我們提供一個交互式計算環境。MATLAB 為我們提供一個交互式計算環境。 在 MATLAB 中可以處理較高層次的數學，例如：處理矩陣代數、傅利級數和其它複雜的函數。

  • FlexTool ( ENM ) 可以存成 m-files 的型式。FlexTool 以模件化, 較有彈性,更容易使用， 開放的環境, 和可靠性為設計重點。FlexTool ( ENM )是以綜合 MATLAB模組的設計。
特性:
  • BUILDING BLOCKS: Upgrade from NN or GA to ENM, upgrade from ENM to CI
  • Robust Neural Networks: Using proprietory Algorithms
  • Off-line learning: Sequencer, Complete System, Partial System, Use NN and EA techniques
  • Niching module: to identify multiple solutions
  • Clustering module: Use separately or with Niching module
  • Optimization: Single and Multiple Objectives
  • High speed evolution : Proprietary Flex-GA algorithm
  • Three Tools in One : Modular, User Friendly, and System Transparent ENM, GA, NN
  • GUI : Easy to use, user friendly
  • Help : Online
  • Tutorial : Hands-on tutorial, application guidelines
  • Parameter Settings : Default parameter settings for the novice
  • General : Statistics, figures, and data collection
  • Cold Start : (start using previously selected parameters)
  • Warm Start : (start from previous generation)
  • learning phase
    1. Off line learning of : neuron associations and weights using genetic algorithms
    2. On line application of : the evolved or known Neural Network
  • GA options : generational, steady state, micro, Flex-GA
  • Coding schemes : include binary, logarithmic (real)
  • Selection : tournament, roulette wheel, ranking
  • Crossover : include 1, 2, multiple point crossover
  • Neural Network Type Options :
    1. BP Network--Fully Connected
    2. BP Recurrent Network
    3. BP Network--Hidden Layers
    4. RBF Network
  • Activation Function Choices :
    1. Linear
    2. Hard Limiter
    3. Piece-wise Linear
    4. Sigmoidal
    5. Bipolar Sigmoidal
    6. Multimodal Sigmoidal
    7. Gaussian
  • Error Computation Options :
    1. Least Square Error
    2. Hampel Error
    3. Huber Error
    4. Logistic Error
    5. Talvars Error
    6. Linear Error
  • Special options include Robust Neural Networks, pattern and batch learning,and validation training.