常用软件包和环境配置（机器学习）

来源：互联网发布：windows 虚拟机下载编辑：程序博客网时间：2024/06/10 06:31

1. 常用软件包：

Theano

Pylearn2

scikit-neuralnetwork

Caffe

Deeplearning4j

Torch

http://deeplearning.net/software_links/

2. 环境配置

Linux: Ubuntu

Eclipse

PyDev

Python

CUDA

GPU: https://developer.nvidia.com/cuda-gpus

3. 神经网络算法 (neural networks)

http://www.maiziedu.com/course/python/373-3811/

12, 14, 15

4. scikit-learn

http://scikit-learn.org/stable/

5. scikit-neuralnetwork

https://github.com/aigamedev/scikit-neuralnetwork

5.1 安装dependencies:

> pip install numpy scipy theano> pip install -e git+https://github.com/lisa-lab/pylearn2.git#egg=Package

> git clone https://github.com/aigamedev/scikit-neuralnetwork.git> cd scikit-neuralnetwork; python setup.py develop

5.2 安装scikit-neuralnetwork

> pip install scikit-neuralnetwork

5.3 测试：

> pip install nose> nosetests -v sknn.tests

如果出现报错： ImportError: No module named dnn

请确保更新theano:

sudo pip install theano --upgrade

5.4 MNIST数据集：

http://yann.lecun.com/exdb/mnist/

5.6 Demo

> python examples/bench_mnist.py (sknn|lasagne)

OS: Linux: Ubuntu 14.04

安装：

1. Pip (Python 2.7.9或以上自带pip）： sudo apt-get install pyton-pip

2. 尝试安装scikit-neuralnetwork: 需要 numpy scipy theano

sudo pip install scikit-neuralnetwork

错误：SystemError: cannot compile "python.h". Perhaps you need to install python-dev|python-devel

解决方法：

sudo apt-get update

sudo apt-get upgrade gcc

sudo apt-get install python2.7-dev

3. 安装numpy, scipy, theano：

pip install numpy scipy theano

sudo pip install numpy scipy theano (1:03:32)

错误：numpy.distutils.system_info.NotFoundError: no lapack/blas resources found

解决方法：sudo apt-get install libblas-dev liblapack-dev libatlas-base-dev gfortran

sudo pip install scipy

sudo pip install Theano

4. 安装Pylearn2

sudo pip install -e git+https://github.com/lisa-lab/pylearn2.git#egg=Package

需要git

sudo apt-get install git

5. 安装scikit-neuralnetwork

> git clone https://github.com/aigamedev/scikit-neuralnetwork.git> cd scikit-neuralnetwork; python setup.py develop

6. 测试

> sudo pip install nose> nosetests -v sknn.tests

需要安装matplotlib: sudo pip install matplotlib

错误：The following package could not be built: freetype, png
解决方法：sudo apt-get install libpng-dev
         sudo apt-get install libjpeg8-dev
         sudo apt-get install libfreetype6-dev

7. 视觉化显示：

> python examples/plot_mlp.py --params activation

8. 在MNIST上测试

> python examples/bench_mnist.py (sknn|lasagne)