{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"#Minimizing finite sums\n",
"\n",
"by Dominik Csiba and Peter Richtarik\n",
"\n",
"The purpose of this lab is to fool around with an efficient randomized algorithm for minimizing finite sums. In particular, we will be solving the L2-regularized ERM (empirical risk minimization) problem, namely \n",
"\n",
"$$\\min_{w\\in \\mathbb{R}^d} \\left\\{ P(w) \\equiv \\frac{1}{n}\\sum_{i=1}^n \\phi_i(X_i^T w) + \\frac{\\lambda}{2}\\|w\\|_2^2 \\right\\},$$\n",
"using Dual-Free SDCA (dfSDCA) [1, 2]. \n",
"\n",
"The meaning of the terms above:\n",
"- $w\\in \\mathbb{R}^d$ is a linear predictor,\n",
"- $X_1,\\dots,X_n \\in \\mathbb{R}^{m\\times d}$ are training examples (in this code we will assume that $m=1$ and hence the examples are simply just vectors),\n",
"- $\\phi_i:\\mathbb{R}^d \\to \\mathbb{R}$ is the loss incurred on example $i$ (label associated with example $i$ is already incorporated in the definition of $\\phi_i$; this why the function $\\phi_i$ has a subscript),\n",
"- $\\lambda>0$ is a regularization parameter (typically, $\\lambda=\\frac{1}{n}$),\n",
"- $P$ is the regularized empirical risk. \n",
"\n",
"\n",
"$\\textbf{Assumption 1.}$ The loss functions $\\phi_i:\\mathbb{R}^m\\mapsto \\mathbb{R}$, $i=1,\\dots,n$, are $\\tfrac{1}{\\gamma}$-smooth. That is, for all $u,v\\in \\mathbb{R}^m$ the following inequality holds:\n",
"$$\\| \\nabla \\phi_i(u) - \\nabla \\phi_i(v) \\|_2 \\leq \\frac{1}{\\gamma}\\|u-v\\|_2.$$\n",
"This is equivalent to requiring that for all $u,v\\in \\mathbb{R}^m$ one has\n",
"$$ \\phi_i(u+v) \\leq \\phi_i(u) + (\\nabla \\phi_i(u))^\\top v + \\frac{1}{2\\gamma}\\|v\\|_2^2.$$\n",
"\n",
"$\\textbf{Assumption 2.}$ The loss functions $\\phi_i:\\mathbb{R}^m \\mapsto \\mathbb{R}$, $i=1,\\dots,n$, are convex. That is, for all $u,v\\in \\mathbb{R}^m$:\n",
"\n",
"$$\\phi(u) + (\\nabla \\phi_i(u))^\\top v \\leq \\phi_i(u+v).$$\n",
"\n",
"The standard SDCA (Stochastic Dual Coordinate Ascent) algorithm, as its name indicates, can be interreted as applying Stochastic/Randomized Coordinate Descent [3] to the Fenchel dual of the ERM problem. The dual problem is naturaly a concave maximization problem, and hence the word \"ascent\" is used instead of \"descent\". However, we have seen that dfSDCA is not motivated nor analyzed via the dual problem, and in this sense, the method is dual-free.\n",
"\n",
"- [1] S. Shalev-Shwartz. SDCA without duality, arXiv:1502.06177, 2015\n",
"- [2] D. Csiba and P. Richtarik. Primal method for ERM with flexible mini-batching schemes and non-convex losses,\n",
"arXiv:1506.02227, 2015\n",
"- [3] P. Richtarik and M. Takac. Iteration complexity of randomized block-coordinate descent methods for minimizing a composite function, Mathematical Programming 144(2):1-38, 2014 (arXiv:1107.2848)\n",
"\n",
"\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 1 - Preliminary stuff\n",
"\n",
"### Step 1.1 - Load a package for plotting \n",
"\n",
"We first load a package for plotting: PyPlot.\n",
"\n",
"### Step 1.2 - Load a package for sampling from a probability distribution\n",
"\n",
"We then include the package probability_tree.jl with functions that will be useful for sampling from a non-uniform distribution from a finite set $[n]:=\\{1, \\dots, n\\}$. In the lecture, we have considered arbitrary samplings, i.e., arbitrary set-valued mappings with values being the subsets of $[n]$. However, the package probability_tree.jl only implements samplings of the form\n",
"\n",
"$$\\mathbf{P}(S_t = \\{i\\}) = p_i,$$\n",
"\n",
"where $p_i>0$ for all $i\\in [n]$ and $\\sum_{i=1}^n p_i = 1$. In particular, the package probability_tree.jl contains two routines:\n",
"- PTCreate - with input being the weights $p$ and output being an array which we call a probability tree, \n",
"- PTSample - with inpit being a probability tree and and output being a random integer $i$ with probability $p_i$.\n",
"\n",
"### Step 1.3 - Load a package for loading data from an external file\n",
"\n",
"Finally, the package load_matricies.jl contains two routines:\n",
"- ReadLIBSVM - reads a dataset in LIBSVM format, inputs the filename, shape and a boolean classification indicating whether the dataset is for regression or classification, and outputs the matrix and labels\n",
"- ReadData - reads a dataset in a standard format, inputs the filename and a boolean classification indicating whether the dataset is for regression or classification, and outputs the matrix and labels\n",
"\n",
"The data has to be stored in the folder \"data\".\n",
"\n",
"### Step 1.4 - Operator overloading\n",
"\n",
"We will also overload the $\\cdot$ operator so that it outputs a scalar as a result of an inner product of a vecor represented as a sparse matrix and a vector represented as an array. These operations will be used often. The default type of the result, without the operator overloading, is a sparse matrix, which would be inconvenient."
]
},
{
"cell_type": "code",
"execution_count": 33,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"dot (generic function with 9 methods)"
]
},
"execution_count": 33,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"using PyPlot\n",
"\n",
"include(\"probability_tree.jl\")\n",
"\n",
"include(\"load_matrices.jl\")\n",
"\n",
"⋅(x::SparseMatrixCSC{Float64, Int64}, y::Array{Float64,1}) = (x'*y)[1]\n",
"⋅(x::SparseMatrixCSC{Float64, Int64}, y::SparseMatrixCSC{Float64, Int64}) = (x'*y)[1]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 2 - Loss functions\n",
"\n",
"We assume that the $n$ examples $X_1,\\dots,X_n \\in \\mathbb{R}^d$ are stored as columns of a $d\\times n$ matrix $X$ (technically, a \"sparse matrix\" in Julia). We also assume that the associated labels $y_1,\\dots,y_n\\in \\mathbb{R}$ are stored as entries of a vector $y\\in \\mathbb{R}^n$ (technically, an \"array\" in Julia).\n",
"\n",
"In particular, we work with the quadratic loss:\n",
"\n",
"$$ \\phi_i(t) = \\frac{1}{2\\gamma} (t - y_i)^2, \\qquad \\nabla \\phi_i(t) = \\frac{1}{\\gamma} (t-y_i), \\qquad \\phi_i(X_i^\\top w) = \\frac{1}{2\\gamma} (X_{i}^\\top w - y_i)^2, \\qquad \\nabla \\phi_i(X_i^\\top w) = \\frac{1}{\\gamma} (X_i^\\top w - y_i).$$\n",
"\n",
"It can be verified that $\\phi_i$ is indeed $\\tfrac{1}{\\gamma}$-smooth and convex. \n",
"\n",
"We now define a function Loss_quadratic whose output will be a touple:\n",
"- function P(w) \n",
"- function g(w,i) (which evaluates $\\nabla \\phi_i(X_i^\\top w)$). \n",
"\n",
"Indeed, functions are first-class citizens in Julia. They can be an output of other functions. This is great, since it will allow us to write a single dfSDCA code which will call generic functions $P$ and $g$. Whenever we change these functions, the code runs with these new functions without any changes. This makes it easy to write flexible optimization code in Julia.\n",
"\n",
"\n",
"### Exercise A: Logistic Regression\n",
"\n",
"Replace the three ??? signs in the Loss_logistic function below by suitable code so that the function works analogously to the Loss_quadratic function. \n",
"\n",
" Hint: The logistic loss is defined as\n",
"\n",
"$$ \\phi_i(t) = \\frac{4}{\\gamma} \\log \\left(1 + e^{y_i t} \\right)$$\n",
"\n",
"and its derivative is\n",
"\n",
"$$\\nabla \\phi_i(t) = \\frac{4 y_i}{\\gamma \\left(1+ e^{-y_i t}\\right)} \\enspace .$$\n",
"\n",
"Again, it can be verified that $\\phi_i$ is indeed $\\tfrac{1}{\\gamma}$-smooth and convex. "
]
},
{
"cell_type": "code",
"execution_count": 34,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"Loss_logistic (generic function with 1 method)"
]
},
"execution_count": 34,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"function Loss_quadratic(X::SparseMatrixCSC, y::Array{Float64,1}, γ::Float64, λ::Float64)\n",
" n = size(X)[2]\n",
" f(w, i) = 1/(2γ)*(X[:,i]⋅w - y[i])^2 # the overloading is used here\n",
" g(w, i) = 1/γ*(X[:,i]⋅w - y[i]) # and here\n",
" P(w) = 1/n*sum([f(w,i) for i=1:n]) + λ/2*w⋅w\n",
" return (g, P)\n",
"end\n",
"\n",
"function Loss_logistic(X::SparseMatrixCSC, y::Array{Float64,1}, γ::Float64, λ::Float64)\n",
" n = size(X)[2]\n",
" f(w, i) = (4/γ) * log(1 + exp(y[i]*X[:,i]⋅w))\n",
" g(w, i) = (4*y[i]) / ( γ * (1 + exp(-y[i]*X[:,i]⋅w)) )\n",
" P(w) = (1/n) * sum([f(w,i) for i=1:n])\n",
" return(g, P)\n",
"end"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 3 - The dfSDCA Algorithm\n",
"\n",
"Recall that the dfSDCA algorithm looks as follows:\n",
"\n",
"Initialize:\n",
"- Choose a sampling (random set-valued mapping) $\\hat{S}$ \n",
"- Compute $p_i = \\mathbf{P}(i \\in \\hat{S})$\n",
"- Compute \"stepsize\" parameters $v_1,\\dots,v_n$ satisfying the ESO inequality\n",
"$$P \\circ X^\\top X \\preceq Diag(p \\circ v),$$\n",
"where $v = (v_1,\\dots,v_n)$, $p=(p_1,\\dots,p_n)$ and $P$ is the $n\\times n$ \"probability\" matrix associated with $\\hat{S}$. That is, $P$ is the matrix with $(i,j)$ entry equal to $\\mathbf{P}(\\{i,j\\}\\subseteq \\hat{S})$. Note that we are now using $P$ both for the probability matrix and the primal objective function. This is too bad, but then no confusion should be caused by this! This form of ESO only makes sense if $m=1$; for $m>1$, the correct ESO inequality looks a bit different [2]. \n",
"- Set \n",
"$$\\theta = \\min_{i} \\frac{p_i \\lambda \\gamma n}{v_i + \\lambda \\gamma n}$$\n",
"- Choose initial \"dual\" variables $\\alpha_1^{0}, \\dots, \\alpha_n^{0} \\in \\mathbb{R}^m$\n",
"- Set \n",
"$$w^{0} = \\frac{1}{\\lambda n}\\sum_{i=1}^n X_i\\alpha_i^{0}$$\n",
"\n",
"For $t \\geq 0$ repeat:\n",
"- Choose a random set (minibatch) $S_t\\subseteq \\{1,2,\\dots,n\\}$ of examples, with $S_t\\sim \\hat{S}$ \n",
"- For $i \\in S_t$ do (possibly in parallel): \n",
"$$\\alpha_i^{t+1} = \\left( 1- \\frac{\\theta}{p_i} \\right) \\alpha_i^{t} + \\frac{\\theta}{p_i}(-\\nabla\\phi_i(X_i^\\top w^{t}))$$\n",
"- Update the main variable:\n",
"$$w^{t+1} = w^{t} - \\sum_{i\\in S_t} \\frac{\\theta}{\\lambda n p_i} X_i(\\nabla\\phi_i(X_i^\\top w^{t}) + \\alpha_i^{t})$$\n",
"\n",
"\n",
"Now we write a Julia function called dfSDCA which implements the algorithm in the special case when $m=1$ and $|\\hat{S}|=1$ with probability 1. It will have the following input variables:\n",
"- $X$ = $d\\times n$ data matrix\n",
"- $g$ = function evaluating $\\phi_i(X_i^\\top w)$\n",
"- $P$ = the objective function\n",
"- $p$ = an array representing a discrete probability distribution over $\\{1,2,\\dots,n \\}$\n",
"- $T$ = number of iterations\n",
"- $\\gamma$ = smoothness parameter (we assume $\\phi_i$ is $1/\\gamma$-smooth)\n",
"- $\\lambda$ = strong-convexity parameter\n",
"- track = a boolean deciding whether we want to output function values or the primal variable $w^{T}$\n",
"- AlgLabel = label for the print outputs during the run of the algorithm\n",
"- progress = a boolean value indicating whether the progress is shown in the console or not\n",
"\n",
"### Exercise B\n",
"\n",
"There are 5 places with questions marks ??? in the dfSDCA code below. Replace them with the correct code."
]
},
{
"cell_type": "code",
"execution_count": 35,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"dfSDCA (generic function with 1 method)"
]
},
"execution_count": 35,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"function dfSDCA(X::SparseMatrixCSC, g::Function, P::Function, p::Array{Float64,1}, T::Int64, \n",
" γ::Float64, λ::Float64, track::Bool, AlgLabel::ASCIIString, progress::Bool)\n",
" \n",
"(d,n) = size(X)\n",
" \n",
"if track\n",
" Pvalue = zeros(ifloor(T/n)+1) # prepares the array of function values\n",
"end\n",
" \n",
"v = [X[:,i]⋅X[:,i] for i=1:n] # list comprehension\n",
"\n",
"θ = minimum(p*λ*γ*n./(v + λ*γ*n)) # step size\n",
"\n",
"PT = PTCreate(p) # creates the probability tree\n",
"\n",
"α = zeros(n) # the starting points are set to zero vectors for simplicity\n",
"w = zeros(d) # w is set to the sum of alphas -> zero\n",
"\n",
"for t=0:T\n",
"\n",
" if t % n == 0 # every n iterations store the function value and prints a message about progress\n",
" if track\n",
" Pvalue[t/n + 1] = P(w)\n",
" if progress\n",
" println(AlgLabel, \": #passes $(t/n), function value: $(Pvalue[t/n + 1])\")\n",
" end\n",
" else\n",
" if progress\n",
" println(AlgLabel, \": #passes $(t/n)\")\n",
" end\n",
" end\n",
" end\n",
"\n",
" i = PTSample(PT) # sample a coordinate according to distribution p\n",
" Δ = g(w,i) + α[i] # store the update as Δ\n",
" α[i] -= (θ*Δ) / p[i]\n",
" w -= (X[:,i]*θ*Δ) / (p[i]*λ*n) # vector - matrix = matrix\n",
" w = vec(w) # matrix -> vector \n",
"end\n",
"\n",
"println( AlgLabel ,\" finished!\")\n",
"\n",
"if track # return function values or w, dependig on track\n",
" return Pvalue\n",
"else\n",
" return w\n",
"end\n",
" \n",
"end"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 4 - Plotting functions\n",
"\n",
"We now write 2 functions which will be used to visualize the results. \n",
"\n",
"Function ShowConvergence takes the following input:\n",
"- Pvalues - a tuple of function value arrays\n",
"- AlgLabels - a tuple of algorithm labels, for legend\n",
"- Pstar - the optimal function value\n",
"\n",
"Function dfSDCACompare is used for comparing the behavior of several variants of dfSDCA in a single plot. The variants differ in the choice of the probability vector $p$. The function first finds the optimal function value (by running the dfSDCA method for 60 passes over data), and then uses the objective function value found in lieu of the true minimum. This is used in order to plot $P(w^t) - P(w^*)$ on the $y$ axis in logarithmic scale.\n",
"\n",
"Input:\n",
"\n",
"- $X$ = data matrix\n",
"- $g$ = function evaluating $\\phi_i(X_i^\\top w)$\n",
"- $P$ = the primal objective function\n",
"- $T$ = number of iterations\n",
"- $\\gamma$ = smoothness parameter\n",
"- $\\lambda$ = strong-convexity parameter\n",
"- $plist$ = a tuple of probability distributions\n",
"- AlgLabel = tuple of labels for the print outputs during the run of the algorithm and for the legend\n",
"- progress = a boolean indicating, whether the progress is shown in the console"
]
},
{
"cell_type": "code",
"execution_count": 36,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"dfSDCACompare (generic function with 1 method)"
]
},
"execution_count": 36,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"function ShowConvergence(Pvalues, AlgLabels, Pstar::Float64) \n",
" ax = axes()\n",
" for i=1:length(Pvalues)\n",
" plt[:plot](0:(length(Pvalues[i])-1), abs(Pvalues[i]-Pstar), \"-\", linewidth=3.0, label=AlgLabels[i])\n",
" end\n",
" legend(loc=\"upper right\")\n",
" ylabel(L\"$P(w^{t}) - P(w^\\star)$\", fontsize=20)\n",
" xlabel(\"passes over data\")\n",
" ax[:set_yscale](\"log\")\n",
" plt[:show]()\n",
"end\n",
"\n",
"function dfSDCACompare(X::SparseMatrixCSC, g::Function, P::Function, \n",
" T::Int64, γ::Float64, λ::Float64, plist, AlgLabel, progress::Bool)\n",
" v = float([X[:,i]⋅X[:,i] for i=1:n])\n",
" p_imp = v + λ*γ*n\n",
" p_imp = p_imp/sum(p_imp)\n",
" wstar = dfSDCA(X, g, P, p_imp, 2T, γ, λ, false, \"Optimal value\", progress)\n",
" Pstar = P(wstar)\n",
" Pvalues = [dfSDCA(X, g, P, plist[i], T, γ, λ, true, AlgLabel[i], progress) for i = 1:length(plist)]\n",
" ShowConvergence(Pvalues, AlgLabel, Pstar)\n",
"end "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 5 - Run the method on a problem with synthetic data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Step 5.1 - We first generate a random problem"
]
},
{
"cell_type": "code",
"execution_count": 37,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"(g,P)"
]
},
"execution_count": 37,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"n = 10000\n",
"d = 100\n",
"X = sprand(d, n, 0.1)\n",
"y = rand(n)\n",
"γ = 1.0\n",
"λ = 1/n\n",
"g, P = Loss_quadratic(X,y,γ,λ)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Step 5.2 - Now we define three different probability distributions\n",
"\n",
"These will correspond to three different versions of the dfSDCA method."
]
},
{
"cell_type": "code",
"execution_count": 38,
"metadata": {
"collapsed": false
},
"outputs": [
{
"data": {
"text/plain": [
"10000-element Array{Float64,1}:\n",
" 2.46661e-5 \n",
" 0.000141786\n",
" 5.87855e-5 \n",
" 9.7088e-5 \n",
" 2.15949e-5 \n",
" 0.000108825\n",
" 9.55093e-5 \n",
" 7.20938e-5 \n",
" 0.00015857 \n",
" 9.44038e-5 \n",
" 0.000168654\n",
" 2.42762e-5 \n",
" 8.28434e-5 \n",
" ⋮ \n",
" 0.000179148\n",
" 0.000102454\n",
" 0.000123156\n",
" 0.000127909\n",
" 9.14772e-5 \n",
" 3.69304e-5 \n",
" 0.000171088\n",
" 5.48171e-5 \n",
" 3.81443e-5 \n",
" 0.000143414\n",
" 5.78489e-5 \n",
" 0.000130837"
]
},
"execution_count": 38,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"p1 = ones(n)/n\n",
"v = float([X[:,i]⋅X[:,i] for i=1:n]) # julia is having difficulties with arrays of type \"Any\"\n",
"p2 = v + n*λ*γ\n",
"p2 = p2/sum(p2)\n",
"p3 = 0.1 + rand(n)\n",
"p3 = p3/sum(p3)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Step 5.3 - We can now run the method"
]
},
{
"cell_type": "code",
"execution_count": 39,
"metadata": {
"collapsed": false
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Optimal value finished!\n",
"dfSDCA-unif finished!\n",
"dfSDCA-imp finished!\n",
"dfSDCA-rand"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
"Figure(PyObject )"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
" finished!\n"
]
}
],
"source": [
"dfSDCACompare(X, g, P, 30n, γ, λ, (p1,p2,p3), (\"dfSDCA-unif\", \"dfSDCA-imp\", \"dfSDCA-rand\"), false)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 6 - Run the method on a problem with real data"
]
},
{
"cell_type": "code",
"execution_count": 40,
"metadata": {
"collapsed": false
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Optimal value: #passes 0.0\n",
"Optimal value: #passes 1.0\n",
"Optimal value: #passes 2.0\n",
"Optimal value: #passes 3.0\n",
"Optimal value: #passes 4.0\n",
"Optimal value: #passes 5.0\n",
"Optimal value: #passes 6.0\n",
"Optimal value: #passes 7.0\n",
"Optimal value: #passes 8.0\n",
"Optimal value: #passes 9.0\n",
"Optimal value: #passes 10.0\n",
"Optimal value: #passes 11.0\n",
"Optimal value: #passes 12.0\n",
"Optimal value: #passes 13.0\n",
"Optimal value: #passes 14.0\n",
"Optimal value: #passes 15.0\n",
"Optimal value: #passes 16.0\n",
"Optimal value: #passes 17.0\n",
"Optimal value: #passes 18.0\n",
"Optimal value: #passes 19.0\n",
"Optimal value: #passes 20.0\n",
"Optimal value: #passes 21.0\n",
"Optimal value: #passes 22.0\n",
"Optimal value: #passes 23.0\n",
"Optimal value: #passes 24.0\n",
"Optimal value: #passes 25.0\n",
"Optimal value: #passes 26.0\n",
"Optimal value: #passes 27.0\n",
"Optimal value: #passes 28.0\n",
"Optimal value: #passes 29.0\n",
"Optimal value: #passes 30.0\n",
"Optimal value: #passes 31.0\n",
"Optimal value: #passes 32.0\n",
"Optimal value: #passes 33.0\n",
"Optimal value: #passes 34.0\n",
"Optimal value: #passes 35.0\n",
"Optimal value: #passes 36.0\n",
"Optimal value: #passes 37.0\n",
"Optimal value: #passes 38.0\n",
"Optimal value: #passes 39.0\n",
"Optimal value: #passes 40.0\n",
"Optimal value: #passes 41.0\n",
"Optimal value: #passes 42.0\n",
"Optimal value: #passes 43.0\n",
"Optimal value: #passes 44.0\n",
"Optimal value: #passes 45.0\n",
"Optimal value: #passes 46.0\n",
"Optimal value: #passes 47.0\n",
"Optimal value: #passes 48.0\n",
"Optimal value: #passes 49.0\n",
"Optimal value: #passes 50.0\n",
"Optimal value: #passes 51.0\n",
"Optimal value: #passes 52.0\n",
"Optimal value: #passes 53.0\n",
"Optimal value: #passes 54.0\n",
"Optimal value: #passes 55.0\n",
"Optimal value: #passes 56.0\n",
"Optimal value: #passes 57.0\n",
"Optimal value: #passes 58.0\n",
"Optimal value: #passes 59.0\n",
"Optimal value: #passes 60.0\n",
"Optimal value finished!\n",
"dfSDCA-unif: #passes 0.0, function value: 2.772588722239787\n",
"dfSDCA-unif: #passes 1.0, function value: 0.043212029486352416\n",
"dfSDCA-unif: #passes 2.0, function value: 0.02071452180356894\n",
"dfSDCA-unif: #passes 3.0, function value: 0.013909429194054095\n",
"dfSDCA-unif: #passes 4.0, function value: 0.011248380824922242\n",
"dfSDCA-unif: #passes 5.0, function value: 0.009965680667403347\n",
"dfSDCA-unif: #passes 6.0, function value: 0.009187388004610202\n",
"dfSDCA-unif: #passes 7.0, function value: 0.00842645311362987\n",
"dfSDCA-unif: #passes 8.0, function value: 0.008168661763742018\n",
"dfSDCA-unif: #passes 9.0, function value: 0.007735400033274704\n",
"dfSDCA-unif: #passes 10.0, function value: 0.0075098141983736535\n",
"dfSDCA-unif: #passes 11.0, function value: 0.0075343004656844475\n",
"dfSDCA-unif: #passes 12.0, function value: 0.007327734205427952\n",
"dfSDCA-unif: #passes 13.0, function value: 0.007247253246714289\n",
"dfSDCA-unif: #passes 14.0, function value: 0.007134328025207796\n",
"dfSDCA-unif: #passes 15.0, function value: 0.007111124545639512\n",
"dfSDCA-unif: #passes 16.0, function value: 0.007036591888370848\n",
"dfSDCA-unif: #passes 17.0, function value: 0.006982013005329268\n",
"dfSDCA-unif: #passes 18.0, function value: 0.006961312552299898\n",
"dfSDCA-unif: #passes 19.0, function value: 0.0069212040905595506\n",
"dfSDCA-unif: #passes 20.0, function value: 0.006924371355776268\n",
"dfSDCA-unif: #passes 21.0, function value: 0.0069212920450419406\n",
"dfSDCA-unif: #passes 22.0, function value: 0.006888624924073018\n",
"dfSDCA-unif: #passes 23.0, function value: 0.00688120423818308\n",
"dfSDCA-unif: #passes 24.0, function value: 0.006881315690110557\n",
"dfSDCA-unif: #passes 25.0, function value: 0.0068566900936668455\n",
"dfSDCA-unif: #passes 26.0, function value: 0.006883086785616939\n",
"dfSDCA-unif: #passes 27.0, function value: 0.0068869982047937905\n",
"dfSDCA-unif: #passes 28.0, function value: 0.006871735222130873\n",
"dfSDCA-unif: #passes 29.0, function value: 0.006875623483277727\n",
"dfSDCA-unif: #passes 30.0, function value: 0.006845505075960709\n",
"dfSDCA-unif finished!\n",
"dfSDCA-imp: #passes 0.0, function value: 2.772588722239787\n",
"dfSDCA-imp: #passes 1.0, function value: 0.03660650635097471\n",
"dfSDCA-imp: #passes 2.0, function value: 0.01965564750439374\n",
"dfSDCA-imp: #passes 3.0, function value: 0.013965801970620103\n",
"dfSDCA-imp: #passes 4.0, function value: 0.011630194703851335\n",
"dfSDCA-imp: #passes 5.0, function value: 0.010261056617613026\n",
"dfSDCA-imp: #passes 6.0, function value: 0.009189104955881683\n",
"dfSDCA-imp: #passes 7.0, function value: 0.008589514207662326\n",
"dfSDCA-imp: #passes 8.0, function value: 0.008266366829997453\n",
"dfSDCA-imp: #passes 9.0, function value: 0.007906625677511301\n",
"dfSDCA-imp: #passes 10.0, function value: 0.007660867730719501\n",
"dfSDCA-imp: #passes 11.0, function value: 0.007520058329473596\n",
"dfSDCA-imp: #passes 12.0, function value: 0.007455858334814755\n",
"dfSDCA-imp: #passes 13.0, function value: 0.0072719507905639105\n",
"dfSDCA-imp: #passes 14.0, function value: 0.007225468190202402\n",
"dfSDCA-imp: #passes 15.0, function value: 0.0072254848581319074\n",
"dfSDCA-imp: #passes 16.0, function value: 0.007135827043393718\n",
"dfSDCA-imp: #passes 17.0, function value: 0.00707711073324589\n",
"dfSDCA-imp: #passes 18.0, function value: 0.00701123287153216\n",
"dfSDCA-imp: #passes 19.0, function value: 0.006984269906035772\n",
"dfSDCA-imp: #passes 20.0, function value: 0.006948982583978348\n",
"dfSDCA-imp: #passes 21.0, function value: 0.006957574739355453\n",
"dfSDCA-imp: #passes 22.0, function value: 0.006935581774228442\n",
"dfSDCA-imp: #passes 23.0, function value: 0.006910705731832703\n",
"dfSDCA-imp: #passes 24.0, function value: 0.006887939749668824\n",
"dfSDCA-imp: #passes 25.0, function value: 0.006866454046152197\n",
"dfSDCA-imp: #passes 26.0, function value: 0.006862355750723067\n",
"dfSDCA-imp: #passes 27.0, function value: 0.006853680680794167\n",
"dfSDCA-imp: #passes 28.0, function value: 0.006855928169347622\n",
"dfSDCA-imp: #passes 29.0, function value: 0.006863378168902077\n",
"dfSDCA-imp: #passes 30.0, function value: 0.006848854685099483\n",
"dfSDCA-imp finished!\n",
"dfSDCA-rand: #passes 0.0, function value: 2.772588722239787\n",
"dfSDCA-rand: #passes 1.0, function value: 0.14076782743476404\n",
"dfSDCA-rand: #passes 2.0, function value: 0.08726708814878126\n",
"dfSDCA-rand: #passes 3.0, function value: 0.062451206403397984\n",
"dfSDCA-rand: #passes 4.0, function value: 0.047876740789458284\n",
"dfSDCA-rand: #passes 5.0, function value: 0.039410271565383154\n",
"dfSDCA-rand: #passes 6.0, function value: 0.0351262172669086\n",
"dfSDCA-rand: #passes 7.0, function value: 0.029419312228186646\n",
"dfSDCA-rand: #passes 8.0, function value: 0.02611978602757744\n",
"dfSDCA-rand: #passes 9.0, function value: 0.023608792653361303\n",
"dfSDCA-rand: #passes 10.0, function value: 0.021524221235268353\n",
"dfSDCA-rand: #passes 11.0, function value: 0.020189559667793986\n",
"dfSDCA-rand: #passes 12.0, function value: 0.01863327596005161\n",
"dfSDCA-rand: #passes 13.0, function value: 0.017579355277910956\n",
"dfSDCA-rand: #passes 14.0, function value: 0.016645476279689028\n",
"dfSDCA-rand: #passes 15.0, function value: 0.015593179924876702\n",
"dfSDCA-rand: #passes 16.0, function value: 0.015000672603990047\n",
"dfSDCA-rand: #passes 17.0, function value: 0.014299815637218714\n",
"dfSDCA-rand: #passes 18.0, function value: 0.013791218003038478\n",
"dfSDCA-rand: #passes 19.0, function value: 0.013255200452482406\n",
"dfSDCA-rand: #passes 20.0, function value: 0.012785709801111643\n",
"dfSDCA-rand: #passes 21.0, function value: 0.012357645925166513\n",
"dfSDCA-rand: #passes 22.0, function value: 0.012026391750105172\n",
"dfSDCA-rand: #passes 23.0, function value: 0.011795380561627202\n",
"dfSDCA-rand: #passes 24.0, function value: 0.011431599625768932\n",
"dfSDCA-rand: #passes 25.0, function value: 0.011107774714348137\n",
"dfSDCA-rand: #passes 26.0, function value: 0.010903307522437436\n",
"dfSDCA-rand: #passes 27.0, function value: 0.010599996253933246\n",
"dfSDCA-rand: #passes 28.0, function value: 0.010463853558015905\n",
"dfSDCA-rand: #passes 29.0, function value: 0.010176451378736686\n",
"dfSDCA-rand"
]
},
{
"data": {
"image/png": "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",
"text/plain": [
"Figure(PyObject )"
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
": #passes 30.0, function value: 0.009984499570298155\n",
"dfSDCA-rand finished!\n"
]
}
],
"source": [
"X,y = ReadData(\"mushrooms\", true)\n",
"#X,y = ReadLIBSVM(\"ijcnn1.tr\", (49990,22) ,true)\n",
"\n",
"d,n = size(X)\n",
"γ = 1.0\n",
"λ = 1/n\n",
"g, P = Loss_logistic(X,y,γ,λ)\n",
"\n",
"p1 = ones(n)/n\n",
"v = float([X[:,i]⋅X[:,i] for i=1:n])\n",
"p2 = v + λ*γ*n\n",
"p2 = p2/sum(p2)\n",
"p3 = 0.1 + rand(n)\n",
"p3 = p3/sum(p3)\n",
"\n",
"dfSDCACompare(X, g, P, 30n, γ, λ, (p1,p2,p3), (\"dfSDCA-unif\", \"dfSDCA-imp\", \"dfSDCA-rand\"), true)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"## Step 7 - Further Exercises\n",
"\n",
"### Exercise C\n",
"\n",
"Write a new function, dfSDCA10, which always updates a set of $10$ variables $\\alpha_i$, chosen uniformly at random. That is, $|S_t|=10$ with probability 1, and all such subsets of cardinality 10 are equaly likely chosen. Make sure the stepsize parameters $v$ are chosen correctly, following theoretical recommendation.\n",
"\n",
"### Exercise D\n",
"\n",
"Write a new function, dfSDCAall, which in each iteration updates all variables $\\alpha_i$. This method is deterministic. Compare its behavior with dfSDCA with uniform sampling of a single \"dual\" variable in a single plot.\n",
"\n",
"### Exercise E\n",
"\n",
"Code up the stochastic gradient descent algorithm (with a fixed stepsize). For a range of stepsizes, plot its behavior (as it is done for dfSDCA in the above plots). Compare with dfSDCA.\n",
"\n",
"### Exercise F\n",
"\n",
"Code up any (or all) of the special instances of the general randomized method for solving linear systems we discussed during the course:\n",
"- Randomized Kaczmarz\n",
"- Randomized Coordinate Descent\n",
"- Randomized Newton\n",
"- Randomized Gaussian Descent\n",
"\n",
"Generate some random problems and play with the solver(s).\n",
"\n",
"### *Exercise G\n",
"\n",
"Find a convex quadratic function $f: \\mathbb{R}^3\\mapsto \\mathbb{R}$ for which the Randomized Newton method with $|C|=2$ converges 1000 times faster than Randomized Newton with $|C|=1$. By convergence we mean $f(x^t)-f(x^*)\\leq 10^{-6}$.\n",
"\n",
"### *Exercise H\n",
"\n",
"Does it make sense to update the probability vector $p$ throughout the iterations of dfSDCA? Can you come up with an adaptive strategy that leads to a faster method in runtime? "
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Julia 0.3.11",
"language": "julia",
"name": "julia-0.3"
},
"language_info": {
"name": "julia",
"version": "0.3.11"
}
},
"nbformat": 4,
"nbformat_minor": 0
}