JNI Should I call DeleteLocalRef

In JNI, FindClass method returns a local reference.
Sample code:
 jclass arrayListClass = env->FindClass("java/util/ArrayList");
Is it necessary to call DeleteLocalRef ? In fact, it is unnecessary to call DeleteLocalRef in most of the cases. However, JNI has a limited (but configurable) number of local references available. So, the best practice is to delete if the reference is created in a loop


C++ Version header template

Make a note for the usage of my version.hpp.
I can not only get the string of version from LIB_VERSION but also get major/minor/patch versions separately.
#pragma once
//  A *string*, LIB_VERSION, in the form "x.y.[z]"
//  where x is the major version number,
//  y is the minor version number,
//  and z is the patch level

#define LIB_VERSION_MAJOR    0
#define LIB_VERSION_MINOR    4
#define LIB_VERSION_PATCH    12

#define AUX_STR_EXP(__A)     #__A
#define AUX_STR(__A)         AUX_STR_EXP(__A)


Besides, I wrote a python script to update the patch.

def changeSoVersion(versionFile):
    # Read in the file
    filedata = None
    with open(versionFile, 'r') as file:
        filedata = file.read()
        matches = re.finditer(REG_VERSION_LINE, filedata)
        for matchNum, match in enumerate(matches):
            oldVersion = match.group(1)
            newVersion = str(int(match.group(1)) + 1)
            print oldVersion
            print newVersion
            if len(oldVersion) > 0 and len(newVersion) > 0:
                filedata = filedata.replace(oldVersion, newVersion)

    # Write the file out again
    with open(versionFile, 'w') as file:


什麼是 closed-form solution and numerical solution

最近正在學解決multivariate analysis和optimalization問題, 遇到一個英文名詞close-form solution. close-form solution中文文獻稱閉合解. 白來來說就是給予一些觀察到的資料, 然後問題可以用函數和數學運算來表示, 這樣的方程式就是close-form solution or numerical solution, 差別在於一個是exact, 一個是approximate. 舉例來說, 在Linear regression中的 Least square equation就是close-form solution, Non-linear regression是numerical solution.