Machine Learning Python Package

Source: Internet
Author: User

With the daily increase in machine learning, a variety of related open source package is also layered out of the face of so many kinds of toolkit, how to choose, and some even know little or do not know, this article simply summarizes the current use of more Python version of the Machine Learning Toolkit, for everyone to see, is not exhaustive, will continue to update, but also welcome you to add, reply! ~~~

Scikit-learn:

A python -based machine learning module based on BSD open source licenses. The project was first initiated by Davidcournapeau in 2007 and is currently being maintained by community volunteers. Basic functions are mainly divided into six parts, classification, regression, clustering, data reduction, model selection, data preprocessing, in particular, can refer to the official website of the document.

Theano:

Debugging process laborious, you can write code, but the derivation and GPU transparent, more optimistic.

Most of the code is written using Cython, and Cython is a Python dialect that can be compiled into native executable code, which can speed up the run faster than using only the explanatory Python language. Most importantly, many optimizations have been integrated into the Theano library, which optimizes your computing and keeps your uptime to a minimum.

It also has built-in support for using CUDA to perform all the time-consuming computations on the GPU. All of this just needs to modify the flag bits in the configuration file. Run a script on the CPU and switch to the GPU, and for your code, you don't need to make any changes.

Although Theano uses Cython and cuda to improve its performance, you can still use only the Python language to create almost any type of neural network structure.

Numbapro:

It's cumbersome to write your own GPU acceleration, and it takes time to write some basic function writing.

Caffe:

A clear and efficient deep learning framework, Caffe is a pure C++/cuda architecture that supports command line, Python, and MATLAB interfaces, and can be seamlessly switched directly between the CPU and the GPU.

Advantages of Caffe:

(1) Quick start: models and corresponding optimizations are given in textual form rather than code. Caffe gives the model definition, optimization settings and pre-training weights for immediate access.

(2) Fast: Able to run the best models and massive amounts of data. Using Caffe with CUDNN, test the alexnet model, and process each picture on K40 with only 1.17ms.

(3) Modularity: Easy to extend to new tasks and settings. You can use the various layer types provided by Caffe to define your own model.

(4) Openness: Open code and reference models are used for reproduction.

(5) Good community: can participate in development and discussion through BSD-2.

pylearn2:

Not flexible.

PYLEARN2 and Theano are developed by the same development team, PYLEARN2 is a machine learning library that encapsulates deep learning and AI research with many common models and training algorithms into a single experimental package, such as a random gradient descent.

You can also easily write a wrapper around your class and algorithm, in order for it to run on Pylearn2, you need to configure the parameters of your entire neural network model in a separate YAML format configuration file.

Nolearn + lasagne:

Kaggle are commonly used by lasagne, and it is now said that the best

Lasagne is a deep learning library with similar functions to blocks and keras, but it is somewhat different in design.

Here are some of lasagne's design objectives:

(1) Simplification: it should be an easy-to-use and extensible machine Learning Library. For each feature you add, you should consider its impact on ease of use and extensibility. The addition of each abstract concept should be carefully examined to determine whether the added complexity is justified.

(2) Small interface: As few classes and methods as possible. Depending on the Theano function and data type as much as possible, follow the Theano rules. If it is not strictly necessary, do not encapsulate things in the class. This makes it easier to use the library and extends it (without much cognition).

(3) Not in the way: unused features should not be visible and users will not consider the features they do not use. Use the components in the library file as separate as possible.

(4) Transparency: Do not attempt to cover up Theano, try to return functions and methods to the Theano expression in the form of a python or numpy data type.

(5) Focus: Follow the Unix philosophy "do one thing and do it well", focusing on feedforward neural networks.

(6) Pragmatism: Making a common use case easier to use is more important than supporting every possible use case.

Keras:

is a simple, highly modular neural network library designed to reference torch, written in Theano and Python languages, with GPU and CPU support. Its development focuses on the implementation of rapid experimentation and the creation of new deep learning models.

If you need a deep learning library with the following features, using Keras is just right:

(1) prototypes can be easily and quickly built (through overall modularity, very simplified and scalable).

(2) Support convolutional networks and recursive networks, and a combination of both.

(3) Support any connection mode (including multi-input multi-output training)

(4) The coding style is very simple and clear. It encapsulates all the essentials in small classes, and it's easy to combine and create a whole new model.

Torch

Temporarily do not understand, welcome to add ~ ~ ~

Blocks:

(1) A very modular framework that helps you build neural networks on Theano.

(2) The features it supports and provides are:

(3) Constructing parameterized Theano operation, called "bricks".

(4) Use pattern matching to select variables and "bricks" in large models.

(5) Use the algorithm to optimize the model.

(6) Save and restore the training model.

(7) Detection and analysis of values during training (training set and test set).

(8) Application of graphic transformations, such as dropout.

Reference: Http://www.csdn.NET/article/2015-08-01/2825362?reload=1

Machine Learning Python Package

Contact Us

The content source of this page is from Internet, which doesn't represent Alibaba Cloud's opinion; products and services mentioned on that page don't have any relationship with Alibaba Cloud. If the content of the page makes you feel confusing, please write us an email, we will handle the problem within 5 days after receiving your email.

If you find any instances of plagiarism from the community, please send an email to: info-contact@alibabacloud.com and provide relevant evidence. A staff member will contact you within 5 working days.

A Free Trial That Lets You Build Big!

Start building with 50+ products and up to 12 months usage for Elastic Compute Service

  • Sales Support

    1 on 1 presale consultation

  • After-Sales Support

    24/7 Technical Support 6 Free Tickets per Quarter Faster Response

  • Alibaba Cloud offers highly flexible support services tailored to meet your exact needs.