随机器学习兴起的Julia编程语言
来源:互联网 发布:中金数据系统 编辑:程序博客网 时间:2024/05/01 19:03
Julia这个编程语言即有Python的开发效率,也有C的执行效率,是为数值运算设计的编程语言。Julia可以直接调用C,很多开源的C和Fortran库都集成到了Julia基础库。另外,它也有notebook。
Julia试图取代R, MATLAB, Octave等数值计算工具。其语法与其他科学计算语言相似。在许多情况下拥有能与编译型语言相媲美的性能。Julia的设计遵从三个原则,快、表达式丰富、动态语言。Julia的核心使用C语言编写,其他部分使用Julia本身编写。
目前这门编程语言在国内知名度不高,如果你在百度搜索Julia,第一页没有一个和Julia语言相关的条目,相反出现的是一个日本av star,这。。。
目前在机器学习领域最流行的编程语言还是Python,看一张图:
一个编程语言的兴衰和背后的社区有直接关系。如果一个编程语言社区强大,那么资源就多,各种库也多,那么用的人就多。Julia的社区貌似都是搞数值运算的,它的应用目前也只限制在这了,如果拿这个语言做Web(有一个库),那不累死。
本帖使用Julia演示一个手写数字识别,看看它的语法是否能和你对上眼。
Julia的几个机器学习库
- ScikitLearn.jl:类似Python的scikit-learn
- Mocha.jl
- TextAnalysis.jl
- MXNet.jl
- TensorFlow.jl:封装TensorFLow
安装Julia
- julia源代码:https://github.com/JuliaLang/julia
- http://julialang.org/downloads/
-
- # Ubuntu
- $ sudo apt install gfortran
- $ sudo apt install julia
- # macOS
- $ brew install Caskroom/cask/julia
- 文档:http://docs.julialang.org/en/stable/manual/
手写数字识别
安装Mocha.jl:
- julia> Pkg.add("Mocha")
- # 或安装最新版 Pkg.clone("https://github.com/pluskid/Mocha.jl.git")
测试安装:准备手写数字数据集:https://github.com/pluskid/Mocha.jl/tree/master/examples/mnist
代码:
- # https://github.com/pluskid/Mocha.jl/blob/master/examples/mnist/mnist.jl
-
- using Mocha
- srand(12345678)
-
- data_layer = AsyncHDF5DataLayer(name="train-data", source="data/train.txt", batch_size=64, shuffle=true)
- conv_layer = ConvolutionLayer(name="conv1", n_filter=20, kernel=(5,5), bottoms=[:data], tops=[:conv])
- pool_layer = PoolingLayer(name="pool1", kernel=(2,2), stride=(2,2), bottoms=[:conv], tops=[:pool])
- conv2_layer = ConvolutionLayer(name="conv2", n_filter=50, kernel=(5,5), bottoms=[:pool], tops=[:conv2])
- pool2_layer = PoolingLayer(name="pool2", kernel=(2,2), stride=(2,2), bottoms=[:conv2], tops=[:pool2])
- fc1_layer = InnerProductLayer(name="ip1", output_dim=500, neuron=Neurons.ReLU(), bottoms=[:pool2], tops=[:ip1])
- fc2_layer = InnerProductLayer(name="ip2", output_dim=10, bottoms=[:ip1], tops=[:ip2])
- loss_layer = SoftmaxLossLayer(name="loss", bottoms=[:ip2,:label])
-
- backend = DefaultBackend()
- init(backend)
-
- common_layers = [conv_layer, pool_layer, conv2_layer, pool2_layer, fc1_layer, fc2_layer]
- net = Net("MNIST-train", backend, [data_layer, common_layers..., loss_layer])
-
- exp_dir = "snapshots-$(Mocha.default_backend_type)"
-
- method = SGD()
- params = make_solver_parameters(method, max_iter=10000, regu_coef=0.0005,
- mom_policy=MomPolicy.Fixed(0.9),
- lr_policy=LRPolicy.Inv(0.01, 0.0001, 0.75),
- load_from=exp_dir)
- solver = Solver(method, params)
-
- setup_coffee_lounge(solver, save_into="$exp_dir/statistics.jld", every_n_iter=1000)
-
- # report training progress every 100 iterations
- add_coffee_break(solver, TrainingSummary(), every_n_iter=100)
-
- # save snapshots every 5000 iterations
- add_coffee_break(solver, Snapshot(exp_dir), every_n_iter=5000)
-
- # show performance on test data every 1000 iterations
- data_layer_test = HDF5DataLayer(name="test-data", source="data/test.txt", batch_size=100)
- acc_layer = AccuracyLayer(name="test-accuracy", bottoms=[:ip2, :label])
- test_net = Net("MNIST-test", backend, [data_layer_test, common_layers..., acc_layer])
- add_coffee_break(solver, ValidationPerformance(test_net), every_n_iter=1000)
-
- solve(solver, net)
-
- #Profile.init(int(1e8), 0.001)
- #@profile solve(solver, net)
- #open("profile.txt", "w") do out
- # Profile.print(out)
- #end
-
- destroy(net)
- destroy(test_net)
- shutdown(backend)