A Brief Study on Machine Learning Tools

A Brief Study on Machine Learning Tools

J. Arockia Jeyanthi

Research Scholar, Department of Computer Science,

St.Xaviers college (Autonomous and Affiliated to Manonmaniam Sundaranar University, Tirunelveli), Palayamkottai, Tamil Nadu, India.

Dr. S. Chidambaranathan

Head and Associate Professor, Department of Computer

Applications, St.Xaviers college (Autonomous and Affiliated to Manonmaniam Sundaranar

University, Tirunelveli), Palayamkottai, Tamil Nadu, India.

Abstract: Machine learning (ML) which is an application of artificial intelligence (AI) provides the systemthe abilityto automatically learn and improve from experience without being programmedexplicitly. ML focuses on the development of computer programs that can access data and use it to learn for themselves. As per the saying The best trained soldiers cant fulfill their mission empty-handed , data scientists are giventheir own weapons known as ML tools. These tools help the data scientiststo deliver results with ease in a machine learning project. Tools are a big part of machine learning and selecting the right tool can be as important as working with the best algorithms. The objective of this paper is to take you to a closer look at popularly used machine learning tools and discover their importance and enable you to choose the types of tools among many.

Keywords: Machine Learning,Tools,Types, Right tool .

1. INTRODUCTION

   Machine Learning (ML) has emerged as the most important technology of the 21st century. For building your very own machine learning model you will take a look at some of the highly popular software among the many prolific algorithms that can be used for designing machine learning solutions.The learning process starts with observations or data, such as examples, direct experience, or instruction, so as to look for patterns in data and make enhanced decisions in the future based on the examples that we give.The main aim is to allow the systems to learn by iteslf without human intervention or support and adjust the actions accordingly. A big part of machine learning are tools and choosing the right tool can be as important as working with the best algorithms.In this paper you will take

   a closer look at machine learning tools and discover their importance and their types that you could select from.

2. WHY USE TOOLS

   The need for the usage of ML tools are described below.

   • Faster: In the applied machine learning process,good tools can automate every step. This means that the time taken from ideas to results is criticallyreduced. The other approach is that you have to implement each capability yourself from scratch. This can take much time than selecting a tool to use off the shelf.
   • Easier:Instead of researching and implementing techniques you can spend your time choosing the good tools. The otherapproach is that you have to be an expert in each step of the process in order to implement it. To ensure it is implemented efficientlyit requires research, deeper exercise in order to understand the techniques, and

     a higher level of engineering.

   • Fun: For beginners to get good resultsthere is a lower barrier. To get better results or work on more projects you can use the extra time. On the other hand rather than on getting resultsyou will spend most of your time building your tools.
3. TOOLS WITH A PURPOSE

   ML tools must serve a solid purpose.These tools facilities to deliver results in a machine learning project. ML tools are not just implementations ofML algorithms.At any step in the process of working through a machine learning problemthe tools provide capabilities that you can use.

4. WHEN TO USE MACHINE LEARNING TOOLS

   Machine learning tools helps you to save time and enable consistency to deliver good results across projects. Some examples of using machine learning tools include:

   • Getting Starting: At your start machine learning tools guide you through the process of delivering worthy results quickly and make you confident to carry on on with your following project.
   • Day-to-Day: Machine learning tools can allow you to focus on the specifics of your problem quicklyrather than on the depths of the techniques you need to use to get an answer.
   • Project Work: For a large project, these tools helps you to prototype a solution, figure out the requirements and provide you a template for the system that you may want to implement.
5. 10+ MOST POPULAR MACHINE LEARNING SOFTWARE TOOLS

   Among the several Machine Learning Softwares available in the market the most popular ones among them are enlisted below.

   1. Scikit-learn :

      It is a free machine learning library for Python. It providessupport to Python numerical and scientific libraries like NumPy and SciPy.

      Features:

      • It aids in data mining and data analysis.
      • It facilitates with models and algorithms for Classification, Regression, Clustering, Dimensional reduction, Model selection, and Pre- processing.
   2. PYTORCH :

      Pytorch was developed by Facebook.It provides an advanced deep learning framework.

      You can develop rapid prototyping for research and can build software pipelines using Pytorch. Using Pytorch , Ubers very own probabilistic programming language is built which is useful to develop dynamic graphs so as to accelerate your machine learning processes.Also it provides your code the ability of data parallelism

      Features:

      • Through Autograd Module it aids in building neural networks.
      • For building neural networks it provides a variety of optimization algorithms.
      • It finds its usage on cloud platforms.
      • It offers various tools,libraries and distributed training.
   3. TENSORFLOW :

      TensorFlow is the standard name for Machine Learning in the Data Science industry. Through its extensive interface of CUDA GPUs it facilitates building of both statistical Machine Learning solutions as well as deep learning. A tensor which is a multi-dimensional array is the most basic data type of TensorFlow.

      It is an open-source toolkit which finds its usage to build machine learning pipelines so that you can build scalable systems to process data.For various applications of ML such as Computer Vision, NLP and Reinforcement Learning ,it provides support and functions. One of the must-know tools of Machine Learning for beginners is TensorFlow.

      Features:

      • It facilitates in training and building your models.
      • With the help of TensorFlow.js which is a model converteryou can run your existing models.
      • It finds its use in the neural network.
   4. WEKA :

      Weka is the acronym for Waikato Environment for Knowledge Analysis. It is a ML software written in Java. It consists of various ML algorithms that can be deployed and are ready for use. These algorithms finds its usage in data mining.

      Weka being an open-source GUI interface allows easy implementation of machine learning algorithms with minimal programming lines.Without writing any line of code you can perform the functioning of machine learning on the data. Hence, it is most suitable for freshers in machine learning.

      Features:

      • Data preparation
      • Classification
      • Regression
      • Clustering
      • Visualization and
      • Association rules mining.</>
   5. KNIME :

      KNIME is the acronym for Konstanz Information Miner .It is an open-source data analytics, reporting as well as an integration platform. One can carry out the various components of machine learning and data mining with the help of KNIME. It is intuitive and is continuously integrating new development features to it. It helps the users to understand the data and design the data science workflows using reusable components which are accessible to all.

      Knime makes use of a modular data pipelining concept. It can blend several data sources to carry out data modeling, analysis, and visualization without the need for extensive programming with the help of GUI and JDBC.

      Features:

      • The code of programming languages like

        C, C++, R, Python, Java, JavaScript etc can be integrated using Knime.

      • It is helpful for financial data analysis,business intelligence and CRM.
   6. COLAB :

      Google Colab which is a cloud service supports Python.Using the libraries of PyTorch, Keras, TensorFlow, and OpenCV it will help you in building the machine learning applications.

      Features:

      • It aids in machine learning education.
      • Helps in machine learning research.
   7. APACHE MAHOUT:

      Apache Mahout is an open-source Machine Learning focused on collaborative filtering as well as classification. These implementations are an extension of the Apache Hadoop Platform. While it is still in progress, the number of algorithms that are supported by it have been growing significantly. Since it is implemented on top of Hadoop, it makes use of the Map/Reduce paradigms.

      Features:

      • It facilitates with algorithms for Pre- processors,Clustering, Recommenders,Regression, and Distributed Linear Algebra.
      • For common math operations, Java libraries are included.
      • Distributed linear algebra framework is followed.

        8) ACCORD.NET :

        For image and audio processing, Accord.Net provides machine learning libraries.

        Features:

        It provides algorithms for:

      • Numerical linear algebra.
      • Numerical optimization
      • Statistics
      • Artificial Neural networks.
      • Image, audio, & signal processing.
      • It also extends support for graph plotting & visualization libraries.
        1. SHOGUN :

           Shogun is a popular, open-source machine learning software.

           Features:

      • It provides support vector machines for regression and classification.
      • It aids in implementing Hidden Markov models.
      • It gives support for many languages like Octave, R, Ruby, Python, Java, Scala, and Lua.
        1. KERAS :

           Keras that provides support for Python is an open-source neural network library. It is well known for its speed, modularity and ease of use. It provides user-friendliness that enables the users to readily implement neural networks without dwelling over the technical jargon when compared to more widely popular libraries like TensorFlow and Pytorch, Keras.

           Features:

      • It can be used for easy and fast prototyping.
      • It provides support to convolution networks.
      • It aids recurrent networks.
      • It provides support to a combination of two networks.
      • It can be run on the CPU and GPU.
        1. RAPID MINER:

   Rapid Miner provides a comprehensive and integrated environment for carrying out several tasks like data preparation, machine learning, text mining, deep learning as well as predictive analytics. It is popular for its lightning-fast speed to reduce costs ,drive revenue and avoid risks.

   One of its most significant features is its GUI based drag and drop feature that allows the users to intuitively build data processing workflows which can be selected from over 2000 available nodes. One can also optimize the model performance through bagging, boosting and building the model ensembles apart from building machine learning models.

   Features:

   • It aids in designing and implementing analytical workflows through GUI,.
   • It assists with data preparation.
   • Result Visualization.
   • Model validation and optimization.
   1. COMPARISON CHART

      CUDA

      Optim Module

      nn Module

      Data preparation

      Classification

      Regression

      	Platform	Cost	Written language	in	Algorithms or Features
      Scikit Learn	Linux, Mac	Free.	Python,		Classification
      	OS, Windows		Cython,	C,	Regression
      			C++		Clustering
      					Preprocessing
      					Model Selection
      					Dimensionality reduction.
      PyTorch	Linux, Mac	Free	Python,	C++,	Autograd Module
      	OS,
      	Windows
      TensorFlow	Linux, Mac	Free	Python,	C++,	Provides a library for dataflow
      	OS,		CUDA		programming.
      	Windows
      Weka	Linux, Mac	Free	Java
      	OS,
      	Windows
      	Clustering
      	Visualization
      	Association rules mining
      KNIME	Linux, Mac	Free	Java		Can work with large data volume.
      	OS,				Supports text mining & image

      mining through plugins

      –

      Supports libraries of PyTorch,

      C#

      Classification

      Python

      API for neural networks

      	Windows
      Colab	Cloud Service	Free
      	Keras, TensorFlow, and OpenCV
      Apache	Cross-platform	Free		Java	Preprocessors
      Mahout				Scala	Regression
      					Clustering
      					Recommenders
      					Distributed Linear Algebra.
      Accors.Net	Cross-platform	Free
      	Regression
      	Distribution
      	Clustering
      	Hypothesis Tests &
      	Kernel Methods
      	Image, Audio & Signal. & Vision
      Shogun	Windows	Free		C++	Regression
      	Linux				Classification
      	UNIX				Clustering
      	Mac OS				Support vector machines.
      					Dimensionality reduction
      					Online learning etc.
      Keras.io	Cross-platform	Free
      Rapid Miner	Cross-platform	Free	plan	Java	Data loading & Transformation
      		Small:	$2500		Data preprocessing &
      		per	year.		visualization.
      		Medium:
      		$5000	per
      		year.
      Large: $10000 per year.

   2. CONCLUSION

In this paper, you have explored machine learning and the top machine learning software in a detailed manner.

Depending on your requirement for the algorithm, your expertise level, and the price,you can select the appropriate tool.

You can use the Machine learning library with ease.Except Rapid Miner most of these libraries are free. TensorFlow which is more popular in machine learning except Rapid Miner except Rapid Miner but it has a learning curve. The two popular tools for machine learning – Scikit-learn and PyTorch both support Python programming language. For neural networks – Keras and TensorFlow are good.