Julia 常用包

ref
trend
observer

零.环境

IJulia 是一个以Julia为后端的交互式环境，可以方便的进行交互式编程

• 安装

using Pkg
Pkg.add("IJulia")

• 使用

using IJulia
notebook()

---

一.数据处理

常用的数据处理包包括以下几个方面：

1.基本科学计算

TODO

2.数据I/O

• CSV

• DataFrames

---

二.绘图

TODO

---

三.机器学习

Mocha

Mocha是一个高效的深度学习框架，包含了通用的随机梯度求解器，可以它构建层训练深、浅(卷积）网络。

• 安装
  Pkg.add("Mocha)
• 使用例子

using Mocha

data  = HDF5DataLayer(name="train-data",source="train-data-list.txt",batch_size=64)
conv  = ConvolutionLayer(name="conv1",n_filter=20,kernel=(5,5),bottoms=[:data],tops=[:conv])
pool  = PoolingLayer(name="pool1",kernel=(2,2),stride=(2,2),bottoms=[:conv],tops=[:pool])
conv2 = ConvolutionLayer(name="conv2",n_filter=50,kernel=(5,5),bottoms=[:pool],tops=[:conv2])
pool2 = PoolingLayer(name="pool2",kernel=(2,2),stride=(2,2),bottoms=[:conv2],tops=[:pool2])
fc1   = InnerProductLayer(name="ip1",output_dim=500,neuron=Neurons.ReLU(),bottoms=[:pool2],
                          tops=[:ip1])
fc2   = InnerProductLayer(name="ip2",output_dim=10,bottoms=[:ip1],tops=[:ip2])
loss  = SoftmaxLossLayer(name="loss",bottoms=[:ip2,:label])

backend = DefaultBackend()
init(backend)

common_layers = [conv, pool, conv2, pool2, fc1, fc2]
net = Net("MNIST-train", backend, [data, common_layers..., loss])

exp_dir = "snapshots"
solver_method = SGD()
params = make_solver_parameters(solver_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(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_test = HDF5DataLayer(name="test-data",source="test-data-list.txt",batch_size=100)
accuracy = AccuracyLayer(name="test-accuracy",bottoms=[:ip2, :label])
test_net = Net("MNIST-test", backend, [data_test, common_layers..., accuracy])
add_coffee_break(solver, ValidationPerformance(test_net), every_n_iter=1000)

solve(solver, net)

destroy(net)
destroy(test_net)
shutdown(backend)

2.Flux

flux是一个机器学习工具包，可以实现各种基本模型（如线性回归）到复杂模型（如神经网络）的搭建、优化和使用。

- 安装

Pkg.add("Flux")
# 可选项目 更新和测试
Pkg.update() # Keep your packages up to date
Pkg.test("Flux") # Check things installed correctly

• 使用,简单模型

#定义模型
W = rand(2, 5)
b = rand(2)

predict(x) = W*x .+ b

function loss(x, y)
  ŷ = predict(x)
  sum((y .- ŷ).^2)
end

x, y = rand(5), rand(2) # Dummy data
loss(x, y) # ~ 3

# 求解梯度
using Flux.Tracker

W = param(W)
b = param(b)

gs = Tracker.gradient(() -> loss(x, y), Params([W, b]))

#更新权重
using Flux.Tracker: update!

Δ = gs[W]

# Update the parameter and reset the gradient
update!(W, -0.1Δ)

loss(x, y) # ~ 2.5

3.Tensorflow

tensorfl.jl基于tensorflow开发的julia封装。
- 安装
Pkg.add("TensorFlow")
- 使用
GPU支持
ENV["TF_USE_GPU"] = "1"
Pkg.build("TensorFlow")
简单的例子：

using TensorFlow

sess = TensorFlow.Session()

x = TensorFlow.constant(Float64[1,2])
y = TensorFlow.Variable(Float64[3,4])
z = TensorFlow.placeholder(Float64)

w = exp(x + z + -y)

run(sess, TensorFlow.global_variables_initializer())
res = run(sess, w, Dict(z=>Float64[1,2]))
Base.Test.@test res[1] ≈ exp(-1)

4.MxNet

MNXET(https://github.com/dmlc/MXNet.jl)julia包

• 安装
  Pkg.add("MXNet")
• 使用

using MXNet
#模型定义
mlp = @mx.chain mx.Variable(:data)             =>
  mx.FullyConnected(name=:fc1, num_hidden=128) =>
  mx.Activation(name=:relu1, act_type=:relu)   =>
  mx.FullyConnected(name=:fc2, num_hidden=64)  =>
  mx.Activation(name=:relu2, act_type=:relu)   =>
  mx.FullyConnected(name=:fc3, num_hidden=10)  =>
  mx.SoftmaxOutput(name=:softmax)

# data provider
batch_size = 100
include(Pkg.dir("MXNet", "examples", "mnist", "mnist-data.jl"))
train_provider, eval_provider = get_mnist_providers(batch_size)

# setup model
model = mx.FeedForward(mlp, context=mx.cpu())

# optimization algorithm
# where η is learning rate and μ is momentum
optimizer = mx.SGD(η=0.1, μ=0.9)

#模型训练
# fit parameters
mx.fit(model, optimizer, train_provider, n_epoch=20, eval_data=eval_provider)

#预测
probs = mx.predict(model, eval_provider)

# collect all labels from eval data
labels = reduce(
  vcat,
  copy(mx.get(eval_provider, batch, :softmax_label)) for batch ∈ eval_provider)
# labels are 0...9
labels .= labels .+ 1

# Now we use compute the accuracy
pred = map(i -> indmax(probs[1:10, i]), 1:size(probs, 2))
correct = sum(pred .== labels)
accuracy = 100correct/length(labels)
@printf "Accuracy on eval set: %.2f%%\n" accuracy

5.Scikit

scikitlearn流行的机器学习包julia实现，支持多种机器学习模型。

- 安装
Pkg.add("ScikitLearn")
- 使用的简单例子

using RDatasets: dataset

iris = dataset("datasets", "iris")
#定义数据
# ScikitLearn.jl expects arrays, but DataFrames can also be used - see
# the corresponding section of the manual
X = convert(Array, iris[[:SepalLength, :SepalWidth, :PetalLength, :PetalWidth]])
y = convert(Array, iris[:Species])

#载入逻辑回归模型
using ScikitLearn

# This model requires scikit-learn. See
# http://scikitlearnjl.readthedocs.io/en/latest/models/#installation
@sk_import linear_model: LogisticRegression
Every model's constructor accepts hyperparameters (such as regression strength, whether to fit the intercept, the penalty type, etc.) as keyword arguments. Check out ?LogisticRegression for details.

model = LogisticRegression(fit_intercept=true)
Then we train the model and evaluate its accuracy on the training set:

#训练
fit!(model, X, y)
#预测
accuracy = sum(predict(model, X) .== y) / length(y)
println("accuracy: $accuracy")

> accuracy: 0.96