スキップしてメイン コンテンツに移動

Implement linear SVM

Introduction


Today, I implement liner SVM (support vector machine).
SVM is one of the most strong algorithm of machine learning before becoming popular Deep learning.
I will write detail of Logic of liner SVM in other posts.

# I finished writing theorem of SVM.
Theorem of SVM part1

My computer is windows. OS is windows. Program is written by Python3.
This program is used Interrior point method in Oputimization.

Dataset

I used two dataset.
First, I used data separated by hyper plane.
Second, I used data mixed \(C_1\) and \(C_2\).
First data is distributed such as the following.
enter image description here
I can write separate line of distribution.
Second data is distributed such as the following.
enter image description here
I can not write separate line of distribution.
I will try liner SVM about these data.

Implementation

  • data1
    enter image description here
This line is hyper plane.
It is written like copperplate.
  • data2
    I tried to estimate hyper plane by variable C.
enter image description here
enter image description here
enter image description here
C value cause this line.
The smoller C value is the easier permitting miss classification.
It is difficult for our to confime effect of C.
It is easy to confime effect of C when I use non liner SVM.
I will write non liner SVM another post.

CODE

My code of liner SVM is opend to the github.
My SVM code is here
File used this time is git_SVM_check.py and git_SVM_def.py .
git_SVM_check.py is main file had following code.
if __name__ == '__main__':
git_SVM_def have class of SVM and code of interror point method.
I will write post of nonliner SVM.
If you see next post also ,I am very glad.
Labels:

コメント

コメントを投稿

このブログの人気の投稿

Implementation of Robbins monro

Robbins monro の実装 sorry, this page is Japanese only.   今回はRobbins monro の実装をしてみました。 Robbins monroは確率勾配降下法の学習率を入りテーション回数の逆数で割っていくものです。 使っているprogram言語はpython 3です。osはwindowsです。(macほしい...) アルゴリズム 確率勾配降下方とは目的関数の最適解を求めるアルゴリズムです。目的関数をf(X)とすると、手順は以下のようになっています。 初期学習率$n_0$を決めます。訓練データDを用意します。この訓練データは複数の初期値の集まりです。 訓練データから一つ初期値をランダムに取り出し、これを$x_0$とし、最初の予測値とします。 次の式に現在の予測値$x_0$を代入し、新たな予測値$x_{n+1}$を得ます。$$x_{n+1} = x_{n} - \frac{n_0}{n} grad f(X_n)$$ 収束して入れば4へ、収束していなければ2で得られた値$x{n+1}$を新たに$x_n$としてもう一度2を行う。 訓練データを一周していなければ2へ、一周していれば各初期値から得られた解の中から目的関数を最も小さくするものを選ぶ。　　 実装例 以下の目的関数を最小化させてみましょう。 $$f(x,y) = (x-2)^2 + (y-3)^2 $$ コマンドラインでpythonを実行していきます。 予想通り、（２，３）という解を導き出してくれました。目的関数が簡単だったので、初期値をどの値でとってもばっちり正解にたどり着いてくれました。 CODE 以下にRobbins monroの関数だけ置いておきます。 こちら にすべてのコードを載せています。 def Robbins_monro(function,grad,number_variable_gradient): init_learning_rate = 1.5 stepsize = 1000 init_value = np.array([range(-1000,1020,20) for i in range(number_v...
コメントを投稿
続きを読む

Graph theory in Python

Introduction 日本語 ver Today, I write the networkx module for graph theory in Python. I write about basis of Gragh theory. Please look at this page . This post is not written Graph theory, but I want to share module in Python for Graph theory. Networkx Python3 have Networkx module. If you have python2, you should install by pip pip install networkx Make Graph by Networkx first, import networkx module and make instance. import networkx as nx import matplotlib.pyplot as plt G = nx.Graph() next, add node and edge G.add_node(1) # add Multiple nodes G.add_nodes_from([2,3,4]) G.add_edge(1,2) # add Multiple edges G.add_edges_from([(3,4),(1,2),(4,6)]) next, plot G nx.draw(G) plt.show() networkx have many function. I will write new learning about networkx in this post. Reference https://qiita.com/kzm4269/items/081ff2fdb8a6b0a6112f http://akiniwa.hatenablog.jp/entry/2013/05/12/012459
コメントを投稿
続きを読む

LIMIT and user variable in MysQL

Problem 日本語 ver I want to use this query in PROCEDURE. SET @user_variable = FLOOR(RAND()*10); INSERT INTO  Table_name (columns_name) VALUES (1) FROM Table_name ORDER BY RAND() LIMIT 1 OFFSET @user_variable; This query want to substitute 1 for the @user_variable 'th line of columns_name, but this query does not work. A cause is that LIMIT and OFFSET does not user variable. Solution SET @user_variable = FLOOR(RAND()*10); PREPARE SET_STMT FROM 'INSERT INTO  Table_name (columns_name) VALUES (1) FROM Table_name ORDER BY RAND() LIMIT 1 OFFSET ?;';  EXECUTE SET_STMT USING @user_variable; Conclusion, if you want to use user variable with LIMIT or OFFSET, use PREPARE STATMENT. Reference http://techtipshoge.blogspot.com/2011/10/limit.html
2 件のコメント
続きを読む