资源简介
源码源自mathworks文件交换中心,优于SMOTE的非平衡学习算法。(有意免金币,但无法设置)
代码片段和文件信息
function [out_featuresSyn out_labelsSyn] = ADASYN(in_features in_labels in_beta in_kDensity in_kSMOTE in_featuresAreNormalized)
%this function implements the ADASYN method as proposed in the following
%paper:
%
%[1]: H. He Y. Bai E.A. Garcia and S. Li “ADASYN: Adaptive Synthetic
%Sampling Approach for Imbalanced Learning“ Proc. Int‘l. J. Conf. Neural
%Networks pp. 1322--1328 (2008).
%
%the implementation follows the notation and equation numbers given in
%section 3.1.4 of another paper:
%
%[2]: H. He and E.A. Garcia “Learning from imbalanced data“
%Knowledge and Data Engineering IEEE Transactions on 21 no. 9
%pp. 1263--1284 (2009).
%
%
%the purpose of the ADASYN method is to improve class balance towards
%equally-sized classes for a given input dataset. this is achieved by
%synthetically creating new examples from the minority class via linear
%interpolation between existing minority class samples. this approach is
%known as the SMOTE method cf. section 3.1.3 in [2]. ADASYN is an
%extension of SMOTE creating more examples in the vicinity of the boundary
%between the two classes than in the interior of the minority class.
%cf. the supplied script demo_ADASYN for an example of this.
%
%
% INPUTS:
%----------
%in_features:
%(N \times P) matrix of numerical features. each row is one example each
%column is one feature hence there are N examples with P features each.
%
%in_labels:
%boolean N-vector of labels defining the classes to which the examples in
%in_features belong.
%
%in_beta [default: 1]:
%desired level of balance where 0 means that the size of the minority
%class will not be changed and 1 means that the minority class will be
%ADASYNed to have (approximately due to rounding) the same size as the
%majority class. any value of in_beta between 0 and 1 provides a compromise
%between these two extremes.
%note that in_beta IS NOT the resulting class ratio but a percentage of
%how much class balance is improved in comparison to the given class
%balance! 0 means nothing is improved in comparison to the given class
%balance and 1 means class sizes are perfectly equalized (except for small
%rounding-related deviations).
%
%in_kDensity [default: 5]:
%k for kNN used in ADASYN density estimation i.e. in calculation of the
%\Gamma_i values in eq. (4) of reference [2]. this is the kNN call that
%regards examples from both classes.
%
%in_kSMOTE [default: 5]:
%k for kNN used in subsequent SMOTE-style synthesis of new examples.
%this is the kNN call that regards only examples from the minority class.
%cf. eq. (1) in reference [2].
%
%in_featuresAreNormalized [default: true]:
%boolean indicating whether the features (i.e. the different columns) in
%in_features are already normalized to the same scale or not.
%by default normalized features are assumed as the input i.e. the user is
%expected to apply a normalization method of choice before passing the data
%to the AD
属性 大小 日期 时间 名称
----------- --------- ---------- ----- ----
文件 12334 2019-02-20 15:17 ADASYN\ADASYN.m
文件 3782 2019-02-20 15:24 ADASYN\demo_ADASYN.m
目录 0 2019-02-20 15:25 ADASYN
----------- --------- ---------- ----- ----
16116 3
- 上一篇:数字式相位差测量multisim仿真
- 下一篇:阿伦方差matlab程序
相关资源
- 斯坦福机器学习编程作业machine-learn
- MATLAB与机器学习
- MACHINE_LEARNING_with_NEURAL_NETWORKS_using_MA
- 交替方向乘子法ADMM算法的matlab代码
- 机器学习 : 实用案例解析 mobi格式
- 带操作界面GUI的字母识别-MATLAB程序
- matlab流形学习算法工具包&matlab机器学
- 机器学习工具包spider工具包
- MATLAB与机器学习 李三平 陈建平译 译
- 机器学习Lasso回归重要论文和Matlab代码
- RVM-MATLAB-V1.3.zip
- 机器学习高斯混合模型资料总结含m
- Coursera吴恩达机器学习课程作业资料
- KNN算法训练MNIST和CIFAR数据集
- Matlab Deep learning 2017年新书
- 机器学习与MATLAB代码
- matlab与机器学习代码
- 机器学习课程设计《基于朴素贝叶斯
- usps手写数字数据集
- cifar-10 数据集 MATLAB版本
- 斯坦福大学CS229机器学习完整详细笔记
- MATLAB神经网络43个案例分析源码
- MATLAB神经网络30个案例分析全书+源代
- 斯坦福机器学习公开课CS229讲义作业及
- 机器学习基础,配套代码
- 机器学习:近20种人工神经网络模型m
- 机器学及其matlab实现—从基础到实践
- mnist的mat格式数据
- 《MATLAB神经网络43个案例分析》源代码
- 机器学习基础教程 matlab代码+数据
评论
共有 条评论