JarCasting
JarCasting JarCasting

ShuJu

Framework for Advanced Statistics and Data Sciences

Лицензия

Лицензия
Группа

Группа

com.github.romualdrousseau
Идентификатор

Идентификатор

shuju
Последняя версия

Последняя версия

1.12
Дата

Дата
Тип

Тип

jar
Описание

Описание

ShuJu
Framework for Advanced Statistics and Data Sciences
Ссылка на сайт

Ссылка на сайт

https://github.com/romualdrousseau/shuju
Система контроля версий

Система контроля версий

https://github.com/romualdrousseau/shuju

Скачать shuju

Имя Файла Размер
shuju-1.12.pom
shuju-1.12.jar 139 KB
shuju-1.12-sources.jar 69 KB
shuju-1.12-javadoc.jar 599 KB
Обзор

Как подключить последнюю версию

<!-- https://jarcasting.com/artifacts/com.github.romualdrousseau/shuju/ -->
<dependency>
    <groupId>com.github.romualdrousseau</groupId>
    <artifactId>shuju</artifactId>
    <version>1.12</version>
</dependency>
// https://jarcasting.com/artifacts/com.github.romualdrousseau/shuju/
implementation 'com.github.romualdrousseau:shuju:1.12'
// https://jarcasting.com/artifacts/com.github.romualdrousseau/shuju/
implementation ("com.github.romualdrousseau:shuju:1.12")
'com.github.romualdrousseau:shuju:jar:1.12'
<dependency org="com.github.romualdrousseau" name="shuju" rev="1.12">
  <artifact name="shuju" type="jar" />
</dependency>
@Grapes(
@Grab(group='com.github.romualdrousseau', module='shuju', version='1.12')
)
libraryDependencies += "com.github.romualdrousseau" % "shuju" % "1.12"
[com.github.romualdrousseau/shuju "1.12"]

Зависимости

compile (3)

Идентификатор библиотеки Тип Версия
org.processing : core jar 3.3.7
com.fasterxml.jackson.core : jackson-databind jar 2.10.0.pr1
org.xerial : sqlite-jdbc jar 3.30.1

test (2)

Идентификатор библиотеки Тип Версия
junit : junit jar 4.13.1
org.hamcrest : hamcrest-library jar 1.3

Модули Проекта

Данный проект не имеет модулей.

ShuJu

Build

It is an enterprise grade Java API for real linear algebra, statistic and machine learning algorithms.

Strongly influenced by Numpy and Tensorflow, ShuJu aims to provide a solid mathematical and scientific library for Java. Despite the success of python and various new languages, Java is still largely used in enterprise and mobile solutions because of its ubiquity and foundations in most of professional platforms.

Each implemented algorithm is not trivial to avoid for example calculation issues as rounding. Also a great care is given to optimization and tests.

The core of the linear algebra is (ala Numpy) an implementation of multi dimensional arrays with the notion of universal functions (lamda calculus). The implementation tries to avoid most of copies during tanspose, reshape and views and use packed java arrays to be friendly with CPU vectorization instructions or GPU. Fast computations is ensured by using BLAS subroutines (Experimental).

So far, the following algorithms have been implemented:

Math library

  • Scalar
  • Tensor (Including Vector, Matrix, ...)
  • Linear Algebra algorithms
    • LU
    • QR
    • HouseGHolder
    • Hessenberg
    • Cholesky
    • Gaussian Elimination
    • Linear Solver
    • Eigens
    • PCA
    • SVD

Neural Network

  • Full connected multiple layers (Dense)
  • Convolution layers
  • DropOut layers
  • Batch Normlization Layers
  • Flatten layers
  • Pooling Layers
  • Genetic Layers
  • Gradient descent back propagation
  • Activation functions (LeakyRelu, Linaer, Relu, Sigmoid, Softmax, Tanh)
  • Loss function (Huber, Hinge, MeanSquare, SoftmaxCrossEntropy)
  • Initializers (Glorot, He, Lecun)
  • Regularizers (L1, L2)
  • Optimizers (Adam, AdaDelta, RMSProp, Sgd)
  • Schedulers (Exponential)

Machine Learnings

  • Kmean
  • K-Nearest Neighbor
  • Naive Bayes
  • Simple Linear Regression

NLP

  • Tokenizers (ngram, shingle)
  • Word 2 vect
  • Regular Expression
  • Stop words

CV

  • Image convolution filter
  • Shape extractors (rectangle by hough transformation)

Genetic

  • Pool with selection
  • Mutation interface

DataSet management

With transformation framework such as normalization, standardization ...

Project Documentation

https://romualdrousseau.github.io/ShuJu/

User Guide

https://github.com/RomualdRousseau/ShuJu/wiki

Install Processing examples

  • Set environement variable PROCESSING_USER_LIBRARIES to the Processing User libraries
  • mvn clean install

Manual Deployment

Manually Deploy snapshot

mvn clean deploy

Manually Deploy release

mvn clean site -Dmessage="<commit_log_here>" scm:checkin -P release
mvn scm:update

Automatic Deployment

Deploy snapshot

Every time a change on master is made (push, merge, etc) a deployment on the snapshot repository is made.

Deploy release

  • Edit the pom.xml to set the release version and push
  • Create a release on github will trigger a deployment on the official repo as well as generate the associated javadoc.
  • Edit the pom.xml to set the snapshot version and push

Deploy documentation

mvn clean site deploy -P documentation

Версии библиотеки

Версия
1.12
1.11
1.10
1.9
1.8
1.7
1.6
1.5
1.4
1.3
1.2
1.1
1.0