diff options
| author | Patrick Wendell <pwendell@apache.org> | 2014-05-30 08:55:36 +0000 |
|---|---|---|
| committer | Patrick Wendell <pwendell@apache.org> | 2014-05-30 08:55:36 +0000 |
| commit | 66fb4b11cbae79d9044b2bbf2e53351642a58ff5 (patch) | |
| tree | 2a3238ae425816d2296c2972987e56b3e949ea0b /site/docs/1.0.0/mllib-optimization.html | |
| parent | 3936fd5d223549bfa06bd6eeba6becfc88daee52 (diff) | |
| download | spark-website-66fb4b11cbae79d9044b2bbf2e53351642a58ff5.tar.gz spark-website-66fb4b11cbae79d9044b2bbf2e53351642a58ff5.tar.bz2 spark-website-66fb4b11cbae79d9044b2bbf2e53351642a58ff5.zip | |
Docs for Spark 1.0.0
Diffstat (limited to 'site/docs/1.0.0/mllib-optimization.html')
| -rw-r--r-- | site/docs/1.0.0/mllib-optimization.html | 428 |
1 files changed, 428 insertions, 0 deletions
diff --git a/site/docs/1.0.0/mllib-optimization.html b/site/docs/1.0.0/mllib-optimization.html new file mode 100644 index 000000000..44483f94f --- /dev/null +++ b/site/docs/1.0.0/mllib-optimization.html @@ -0,0 +1,428 @@ +<!DOCTYPE html> +<!--[if lt IE 7]> <html class="no-js lt-ie9 lt-ie8 lt-ie7"> <![endif]--> +<!--[if IE 7]> <html class="no-js lt-ie9 lt-ie8"> <![endif]--> +<!--[if IE 8]> <html class="no-js lt-ie9"> <![endif]--> +<!--[if gt IE 8]><!--> <html class="no-js"> <!--<![endif]--> + <head> + <meta charset="utf-8"> + <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> + <title>Optimization - MLlib - Spark 1.0.0 Documentation</title> + <meta name="description" content=""> + + + + <link rel="stylesheet" href="css/bootstrap.min.css"> + <style> + body { + padding-top: 60px; + padding-bottom: 40px; + } + </style> + <meta name="viewport" content="width=device-width"> + <link rel="stylesheet" href="css/bootstrap-responsive.min.css"> + <link rel="stylesheet" href="css/main.css"> + + <script src="js/vendor/modernizr-2.6.1-respond-1.1.0.min.js"></script> + + <link rel="stylesheet" href="css/pygments-default.css"> + + + + </head> + <body> + <!--[if lt IE 7]> + <p class="chromeframe">You are using an outdated browser. <a href="http://browsehappy.com/">Upgrade your browser today</a> or <a href="http://www.google.com/chromeframe/?redirect=true">install Google Chrome Frame</a> to better experience this site.</p> + <![endif]--> + + <!-- This code is taken from http://twitter.github.com/bootstrap/examples/hero.html --> + + <div class="navbar navbar-fixed-top" id="topbar"> + <div class="navbar-inner"> + <div class="container"> + <div class="brand"><a href="index.html"> + <img src="img/spark-logo-hd.png" style="height:50px;"/></a><span class="version">1.0.0</span> + </div> + <ul class="nav"> + <!--TODO(andyk): Add class="active" attribute to li some how.--> + <li><a href="index.html">Overview</a></li> + + <li class="dropdown"> + <a href="#" class="dropdown-toggle" data-toggle="dropdown">Programming Guides<b class="caret"></b></a> + <ul class="dropdown-menu"> + <li><a href="quick-start.html">Quick Start</a></li> + <li><a href="programming-guide.html">Spark Programming Guide</a></li> + <li class="divider"></li> + <li><a href="streaming-programming-guide.html">Spark Streaming</a></li> + <li><a href="sql-programming-guide.html">Spark SQL</a></li> + <li><a href="mllib-guide.html">MLlib (Machine Learning)</a></li> + <li><a href="graphx-programming-guide.html">GraphX (Graph Processing)</a></li> + <li><a href="bagel-programming-guide.html">Bagel (Pregel on Spark)</a></li> + </ul> + </li> + + <li class="dropdown"> + <a href="#" class="dropdown-toggle" data-toggle="dropdown">API Docs<b class="caret"></b></a> + <ul class="dropdown-menu"> + <li><a href="api/scala/index.html#org.apache.spark.package">Scaladoc</a></li> + <li><a href="api/java/index.html">Javadoc</a></li> + <li><a href="api/python/index.html">Python API</a></li> + </ul> + </li> + + <li class="dropdown"> + <a href="#" class="dropdown-toggle" data-toggle="dropdown">Deploying<b class="caret"></b></a> + <ul class="dropdown-menu"> + <li><a href="cluster-overview.html">Overview</a></li> + <li><a href="submitting-applications.html">Submitting Applications</a></li> + <li class="divider"></li> + <li><a href="ec2-scripts.html">Amazon EC2</a></li> + <li><a href="spark-standalone.html">Standalone Mode</a></li> + <li><a href="running-on-mesos.html">Mesos</a></li> + <li><a href="running-on-yarn.html">YARN</a></li> + </ul> + </li> + + <li class="dropdown"> + <a href="api.html" class="dropdown-toggle" data-toggle="dropdown">More<b class="caret"></b></a> + <ul class="dropdown-menu"> + <li><a href="configuration.html">Configuration</a></li> + <li><a href="monitoring.html">Monitoring</a></li> + <li><a href="tuning.html">Tuning Guide</a></li> + <li><a href="job-scheduling.html">Job Scheduling</a></li> + <li><a href="security.html">Security</a></li> + <li><a href="hardware-provisioning.html">Hardware Provisioning</a></li> + <li><a href="hadoop-third-party-distributions.html">3<sup>rd</sup>-Party Hadoop Distros</a></li> + <li class="divider"></li> + <li><a href="building-with-maven.html">Building Spark with Maven</a></li> + <li><a href="https://cwiki.apache.org/confluence/display/SPARK/Contributing+to+Spark">Contributing to Spark</a></li> + </ul> + </li> + </ul> + <!--<p class="navbar-text pull-right"><span class="version-text">v1.0.0</span></p>--> + </div> + </div> + </div> + + <div class="container" id="content"> + + <h1 class="title"><a href="mllib-guide.html">MLlib</a> - Optimization</h1> + + + <ul id="markdown-toc"> + <li><a href="#mathematical-description">Mathematical description</a> <ul> + <li><a href="#gradient-descent">Gradient descent</a></li> + <li><a href="#stochastic-gradient-descent-sgd">Stochastic gradient descent (SGD)</a></li> + <li><a href="#update-schemes-for-distributed-sgd">Update schemes for distributed SGD</a></li> + <li><a href="#limited-memory-bfgs-l-bfgs">Limited-memory BFGS (L-BFGS)</a></li> + </ul> + </li> + <li><a href="#implementation-in-mllib">Implementation in MLlib</a> <ul> + <li><a href="#gradient-descent-and-stochastic-gradient-descent">Gradient descent and stochastic gradient descent</a></li> + <li><a href="#l-bfgs">L-BFGS</a> <ul> + <li><a href="#developers-note">Developer’s note</a></li> + </ul> + </li> + </ul> + </li> +</ul> + +<p><code>\[ +\newcommand{\R}{\mathbb{R}} +\newcommand{\E}{\mathbb{E}} +\newcommand{\x}{\mathbf{x}} +\newcommand{\y}{\mathbf{y}} +\newcommand{\wv}{\mathbf{w}} +\newcommand{\av}{\mathbf{\alpha}} +\newcommand{\bv}{\mathbf{b}} +\newcommand{\N}{\mathbb{N}} +\newcommand{\id}{\mathbf{I}} +\newcommand{\ind}{\mathbf{1}} +\newcommand{\0}{\mathbf{0}} +\newcommand{\unit}{\mathbf{e}} +\newcommand{\one}{\mathbf{1}} +\newcommand{\zero}{\mathbf{0}} +\]</code></p> + +<h2 id="mathematical-description">Mathematical description</h2> + +<h3 id="gradient-descent">Gradient descent</h3> +<p>The simplest method to solve optimization problems of the form <code>$\min_{\wv \in\R^d} \; f(\wv)$</code> +is <a href="http://en.wikipedia.org/wiki/Gradient_descent">gradient descent</a>. +Such first-order optimization methods (including gradient descent and stochastic variants +thereof) are well-suited for large-scale and distributed computation.</p> + +<p>Gradient descent methods aim to find a local minimum of a function by iteratively taking steps in +the direction of steepest descent, which is the negative of the derivative (called the +<a href="http://en.wikipedia.org/wiki/Gradient">gradient</a>) of the function at the current point, i.e., at +the current parameter value. +If the objective function <code>$f$</code> is not differentiable at all arguments, but still convex, then a +<em>sub-gradient</em> +is the natural generalization of the gradient, and assumes the role of the step direction. +In any case, computing a gradient or sub-gradient of <code>$f$</code> is expensive — it requires a full +pass through the complete dataset, in order to compute the contributions from all loss terms.</p> + +<h3 id="stochastic-gradient-descent-sgd">Stochastic gradient descent (SGD)</h3> +<p>Optimization problems whose objective function <code>$f$</code> is written as a sum are particularly +suitable to be solved using <em>stochastic gradient descent (SGD)</em>. +In our case, for the optimization formulations commonly used in <a href="mllib-classification-regression.html">supervised machine learning</a>, +<code>\begin{equation} + f(\wv) := + \lambda\, R(\wv) + + \frac1n \sum_{i=1}^n L(\wv;\x_i,y_i) + \label{eq:regPrimal} + \ . +\end{equation}</code> +this is especially natural, because the loss is written as an average of the individual losses +coming from each datapoint.</p> + +<p>A stochastic subgradient is a randomized choice of a vector, such that in expectation, we obtain +a true subgradient of the original objective function. +Picking one datapoint <code>$i\in[1..n]$</code> uniformly at random, we obtain a stochastic subgradient of +<code>$\eqref{eq:regPrimal}$</code>, with respect to <code>$\wv$</code> as follows: +<code>\[ +f'_{\wv,i} := L'_{\wv,i} + \lambda\, R'_\wv \ , +\]</code> +where <code>$L'_{\wv,i} \in \R^d$</code> is a subgradient of the part of the loss function determined by the +<code>$i$</code>-th datapoint, that is <code>$L'_{\wv,i} \in \frac{\partial}{\partial \wv} L(\wv;\x_i,y_i)$</code>. +Furthermore, <code>$R'_\wv$</code> is a subgradient of the regularizer <code>$R(\wv)$</code>, i.e. <code>$R'_\wv \in +\frac{\partial}{\partial \wv} R(\wv)$</code>. The term <code>$R'_\wv$</code> does not depend on which random +datapoint is picked. +Clearly, in expectation over the random choice of <code>$i\in[1..n]$</code>, we have that <code>$f'_{\wv,i}$</code> is +a subgradient of the original objective <code>$f$</code>, meaning that <code>$\E\left[f'_{\wv,i}\right] \in +\frac{\partial}{\partial \wv} f(\wv)$</code>.</p> + +<p>Running SGD now simply becomes walking in the direction of the negative stochastic subgradient +<code>$f'_{\wv,i}$</code>, that is +<code>\begin{equation}\label{eq:SGDupdate} +\wv^{(t+1)} := \wv^{(t)} - \gamma \; f'_{\wv,i} \ . +\end{equation}</code> +<strong>Step-size.</strong> +The parameter <code>$\gamma$</code> is the step-size, which in the default implementation is chosen +decreasing with the square root of the iteration counter, i.e. <code>$\gamma := \frac{s}{\sqrt{t}}$</code> +in the <code>$t$</code>-th iteration, with the input parameter <code>$s=$ stepSize</code>. Note that selecting the best +step-size for SGD methods can often be delicate in practice and is a topic of active research.</p> + +<p><strong>Gradients.</strong> +A table of (sub)gradients of the machine learning methods implemented in MLlib, is available in +the <a href="mllib-classification-regression.html">classification and regression</a> section.</p> + +<p><strong>Proximal Updates.</strong> +As an alternative to just use the subgradient <code>$R'(\wv)$</code> of the regularizer in the step +direction, an improved update for some cases can be obtained by using the proximal operator +instead. +For the L1-regularizer, the proximal operator is given by soft thresholding, as implemented in +<a href="api/scala/index.html#org.apache.spark.mllib.optimization.L1Updater">L1Updater</a>.</p> + +<h3 id="update-schemes-for-distributed-sgd">Update schemes for distributed SGD</h3> +<p>The SGD implementation in +<a href="api/scala/index.html#org.apache.spark.mllib.optimization.GradientDescent">GradientDescent</a> uses +a simple (distributed) sampling of the data examples. +We recall that the loss part of the optimization problem <code>$\eqref{eq:regPrimal}$</code> is +<code>$\frac1n \sum_{i=1}^n L(\wv;\x_i,y_i)$</code>, and therefore <code>$\frac1n \sum_{i=1}^n L'_{\wv,i}$</code> would +be the true (sub)gradient. +Since this would require access to the full data set, the parameter <code>miniBatchFraction</code> specifies +which fraction of the full data to use instead. +The average of the gradients over this subset, i.e. +<code>\[ +\frac1{|S|} \sum_{i\in S} L'_{\wv,i} \ , +\]</code> +is a stochastic gradient. Here <code>$S$</code> is the sampled subset of size <code>$|S|=$ miniBatchFraction +$\cdot n$</code>.</p> + +<p>In each iteration, the sampling over the distributed dataset +(<a href="programming-guide.html#resilient-distributed-datasets-rdds">RDD</a>), as well as the +computation of the sum of the partial results from each worker machine is performed by the +standard spark routines.</p> + +<p>If the fraction of points <code>miniBatchFraction</code> is set to 1 (default), then the resulting step in +each iteration is exact (sub)gradient descent. In this case there is no randomness and no +variance in the used step directions. +On the other extreme, if <code>miniBatchFraction</code> is chosen very small, such that only a single point +is sampled, i.e. <code>$|S|=$ miniBatchFraction $\cdot n = 1$</code>, then the algorithm is equivalent to +standard SGD. In that case, the step direction depends from the uniformly random sampling of the +point.</p> + +<h3 id="limited-memory-bfgs-l-bfgs">Limited-memory BFGS (L-BFGS)</h3> +<p><a href="http://en.wikipedia.org/wiki/Limited-memory_BFGS">L-BFGS</a> is an optimization +algorithm in the family of quasi-Newton methods to solve the optimization problems of the form +<code>$\min_{\wv \in\R^d} \; f(\wv)$</code>. The L-BFGS method approximates the objective function locally as a +quadratic without evaluating the second partial derivatives of the objective function to construct the +Hessian matrix. The Hessian matrix is approximated by previous gradient evaluations, so there is no +vertical scalability issue (the number of training features) when computing the Hessian matrix +explicitly in Newton’s method. As a result, L-BFGS often achieves rapider convergence compared with +other first-order optimization. </p> + +<h2 id="implementation-in-mllib">Implementation in MLlib</h2> + +<h3 id="gradient-descent-and-stochastic-gradient-descent">Gradient descent and stochastic gradient descent</h3> +<p>Gradient descent methods including stochastic subgradient descent (SGD) as +included as a low-level primitive in <code>MLlib</code>, upon which various ML algorithms +are developed, see the +<a href="mllib-linear-methods.html">linear methods</a> +section for example.</p> + +<p>The SGD method +<a href="api/scala/index.html#org.apache.spark.mllib.optimization.GradientDescent">GradientDescent.runMiniBatchSGD</a> +has the following parameters:</p> + +<ul> + <li><code>Gradient</code> is a class that computes the stochastic gradient of the function +being optimized, i.e., with respect to a single training example, at the +current parameter value. MLlib includes gradient classes for common loss +functions, e.g., hinge, logistic, least-squares. The gradient class takes as +input a training example, its label, and the current parameter value. </li> + <li><code>Updater</code> is a class that performs the actual gradient descent step, i.e. +updating the weights in each iteration, for a given gradient of the loss part. +The updater is also responsible to perform the update from the regularization +part. MLlib includes updaters for cases without regularization, as well as +L1 and L2 regularizers.</li> + <li><code>stepSize</code> is a scalar value denoting the initial step size for gradient +descent. All updaters in MLlib use a step size at the t-th step equal to +<code>stepSize $/ \sqrt{t}$</code>. </li> + <li><code>numIterations</code> is the number of iterations to run.</li> + <li><code>regParam</code> is the regularization parameter when using L1 or L2 regularization.</li> + <li><code>miniBatchFraction</code> is the fraction of the total data that is sampled in +each iteration, to compute the gradient direction.</li> +</ul> + +<p>Available algorithms for gradient descent:</p> + +<ul> + <li><a href="api/scala/index.html#org.apache.spark.mllib.optimization.GradientDescent">GradientDescent.runMiniBatchSGD</a></li> +</ul> + +<h3 id="l-bfgs">L-BFGS</h3> +<p>L-BFGS is currently only a low-level optimization primitive in <code>MLlib</code>. If you want to use L-BFGS in various +ML algorithms such as Linear Regression, and Logistic Regression, you have to pass the gradient of objective +function, and updater into optimizer yourself instead of using the training APIs like +<a href="api/scala/index.html#org.apache.spark.mllib.classification.LogisticRegressionWithSGD">LogisticRegressionWithSGD</a>. +See the example below. It will be addressed in the next release. </p> + +<p>The L1 regularization by using +<a href="api/scala/index.html#org.apache.spark.mllib.optimization.L1Updater">L1Updater</a> will not work since the +soft-thresholding logic in L1Updater is designed for gradient descent. See the developer’s note.</p> + +<p>The L-BFGS method +<a href="api/scala/index.html#org.apache.spark.mllib.optimization.LBFGS">LBFGS.runLBFGS</a> +has the following parameters:</p> + +<ul> + <li><code>Gradient</code> is a class that computes the gradient of the objective function +being optimized, i.e., with respect to a single training example, at the +current parameter value. MLlib includes gradient classes for common loss +functions, e.g., hinge, logistic, least-squares. The gradient class takes as +input a training example, its label, and the current parameter value. </li> + <li><code>Updater</code> is a class that computes the gradient and loss of objective function +of the regularization part for L-BFGS. MLlib includes updaters for cases without +regularization, as well as L2 regularizer. </li> + <li><code>numCorrections</code> is the number of corrections used in the L-BFGS update. 10 is +recommended.</li> + <li><code>maxNumIterations</code> is the maximal number of iterations that L-BFGS can be run.</li> + <li><code>regParam</code> is the regularization parameter when using regularization.</li> +</ul> + +<p>The <code>return</code> is a tuple containing two elements. The first element is a column matrix +containing weights for every feature, and the second element is an array containing +the loss computed for every iteration.</p> + +<p>Here is an example to train binary logistic regression with L2 regularization using +L-BFGS optimizer. </p> + +<div class="highlight"><pre><code class="scala"><span class="k">import</span> <span class="nn">org.apache.spark.SparkContext</span> +<span class="k">import</span> <span class="nn">org.apache.spark.mllib.evaluation.BinaryClassificationMetrics</span> +<span class="k">import</span> <span class="nn">org.apache.spark.mllib.linalg.Vectors</span> +<span class="k">import</span> <span class="nn">org.apache.spark.mllib.util.MLUtils</span> +<span class="k">import</span> <span class="nn">org.apache.spark.mllib.classification.LogisticRegressionModel</span> + +<span class="k">val</span> <span class="n">data</span> <span class="k">=</span> <span class="nc">MLUtils</span><span class="o">.</span><span class="n">loadLibSVMFile</span><span class="o">(</span><span class="n">sc</span><span class="o">,</span> <span class="s">"mllib/data/sample_libsvm_data.txt"</span><span class="o">)</span> +<span class="k">val</span> <span class="n">numFeatures</span> <span class="k">=</span> <span class="n">data</span><span class="o">.</span><span class="n">take</span><span class="o">(</span><span class="mi">1</span><span class="o">)(</span><span class="mi">0</span><span class="o">).</span><span class="n">features</span><span class="o">.</span><span class="n">size</span> + +<span class="c1">// Split data into training (60%) and test (40%).</span> +<span class="k">val</span> <span class="n">splits</span> <span class="k">=</span> <span class="n">data</span><span class="o">.</span><span class="n">randomSplit</span><span class="o">(</span><span class="nc">Array</span><span class="o">(</span><span class="mf">0.6</span><span class="o">,</span> <span class="mf">0.4</span><span class="o">),</span> <span class="n">seed</span> <span class="k">=</span> <span class="mi">11L</span><span class="o">)</span> + +<span class="c1">// Append 1 into the training data as intercept.</span> +<span class="k">val</span> <span class="n">training</span> <span class="k">=</span> <span class="n">splits</span><span class="o">(</span><span class="mi">0</span><span class="o">).</span><span class="n">map</span><span class="o">(</span><span class="n">x</span> <span class="k">=></span> <span class="o">(</span><span class="n">x</span><span class="o">.</span><span class="n">label</span><span class="o">,</span> <span class="nc">MLUtils</span><span class="o">.</span><span class="n">appendBias</span><span class="o">(</span><span class="n">x</span><span class="o">.</span><span class="n">features</span><span class="o">))).</span><span class="n">cache</span><span class="o">()</span> + +<span class="k">val</span> <span class="n">test</span> <span class="k">=</span> <span class="n">splits</span><span class="o">(</span><span class="mi">1</span><span class="o">)</span> + +<span class="c1">// Run training algorithm to build the model</span> +<span class="k">val</span> <span class="n">numCorrections</span> <span class="k">=</span> <span class="mi">10</span> +<span class="k">val</span> <span class="n">convergenceTol</span> <span class="k">=</span> <span class="mi">1</span><span class="n">e</span><span class="o">-</span><span class="mi">4</span> +<span class="k">val</span> <span class="n">maxNumIterations</span> <span class="k">=</span> <span class="mi">20</span> +<span class="k">val</span> <span class="n">regParam</span> <span class="k">=</span> <span class="mf">0.1</span> +<span class="k">val</span> <span class="n">initialWeightsWithIntercept</span> <span class="k">=</span> <span class="nc">Vectors</span><span class="o">.</span><span class="n">dense</span><span class="o">(</span><span class="k">new</span> <span class="nc">Array</span><span class="o">[</span><span class="kt">Double</span><span class="o">](</span><span class="n">numFeatures</span> <span class="o">+</span> <span class="mi">1</span><span class="o">))</span> + +<span class="k">val</span> <span class="o">(</span><span class="n">weightsWithIntercept</span><span class="o">,</span> <span class="n">loss</span><span class="o">)</span> <span class="k">=</span> <span class="nc">LBFGS</span><span class="o">.</span><span class="n">runLBFGS</span><span class="o">(</span> + <span class="n">training</span><span class="o">,</span> + <span class="k">new</span> <span class="nc">LogisticGradient</span><span class="o">(),</span> + <span class="k">new</span> <span class="nc">SquaredL2Updater</span><span class="o">(),</span> + <span class="n">numCorrections</span><span class="o">,</span> + <span class="n">convergenceTol</span><span class="o">,</span> + <span class="n">maxNumIterations</span><span class="o">,</span> + <span class="n">regParam</span><span class="o">,</span> + <span class="n">initialWeightsWithIntercept</span><span class="o">)</span> + +<span class="k">val</span> <span class="n">model</span> <span class="k">=</span> <span class="k">new</span> <span class="nc">LogisticRegressionModel</span><span class="o">(</span> + <span class="nc">Vectors</span><span class="o">.</span><span class="n">dense</span><span class="o">(</span><span class="n">weightsWithIntercept</span><span class="o">.</span><span class="n">toArray</span><span class="o">.</span><span class="n">slice</span><span class="o">(</span><span class="mi">0</span><span class="o">,</span> <span class="n">weightsWithIntercept</span><span class="o">.</span><span class="n">size</span> <span class="o">-</span> <span class="mi">1</span><span class="o">)),</span> + <span class="n">weightsWithIntercept</span><span class="o">(</span><span class="n">weightsWithIntercept</span><span class="o">.</span><span class="n">size</span> <span class="o">-</span> <span class="mi">1</span><span class="o">))</span> + +<span class="c1">// Clear the default threshold.</span> +<span class="n">model</span><span class="o">.</span><span class="n">clearThreshold</span><span class="o">()</span> + +<span class="c1">// Compute raw scores on the test set.</span> +<span class="k">val</span> <span class="n">scoreAndLabels</span> <span class="k">=</span> <span class="n">test</span><span class="o">.</span><span class="n">map</span> <span class="o">{</span> <span class="n">point</span> <span class="k">=></span> + <span class="k">val</span> <span class="n">score</span> <span class="k">=</span> <span class="n">model</span><span class="o">.</span><span class="n">predict</span><span class="o">(</span><span class="n">point</span><span class="o">.</span><span class="n">features</span><span class="o">)</span> + <span class="o">(</span><span class="n">score</span><span class="o">,</span> <span class="n">point</span><span class="o">.</span><span class="n">label</span><span class="o">)</span> +<span class="o">}</span> + +<span class="c1">// Get evaluation metrics.</span> +<span class="k">val</span> <span class="n">metrics</span> <span class="k">=</span> <span class="k">new</span> <span class="nc">BinaryClassificationMetrics</span><span class="o">(</span><span class="n">scoreAndLabels</span><span class="o">)</span> +<span class="k">val</span> <span class="n">auROC</span> <span class="k">=</span> <span class="n">metrics</span><span class="o">.</span><span class="n">areaUnderROC</span><span class="o">()</span> + +<span class="n">println</span><span class="o">(</span><span class="s">"Loss of each step in training process"</span><span class="o">)</span> +<span class="n">loss</span><span class="o">.</span><span class="n">foreach</span><span class="o">(</span><span class="n">println</span><span class="o">)</span> +<span class="n">println</span><span class="o">(</span><span class="s">"Area under ROC = "</span> <span class="o">+</span> <span class="n">auROC</span><span class="o">)</span> +</code></pre></div> + +<h4 id="developers-note">Developer’s note</h4> +<p>Since the Hessian is constructed approximately from previous gradient evaluations, +the objective function can not be changed during the optimization process. +As a result, Stochastic L-BFGS will not work naively by just using miniBatch; +therefore, we don’t provide this until we have better understanding.</p> + +<ul> + <li><code>Updater</code> is a class originally designed for gradient decent which computes +the actual gradient descent step. However, we’re able to take the gradient and +loss of objective function of regularization for L-BFGS by ignoring the part of logic +only for gradient decent such as adaptive step size stuff. We will refactorize +this into regularizer to replace updater to separate the logic between +regularization and step update later. </li> +</ul> + + + </div> <!-- /container --> + + <script src="js/vendor/jquery-1.8.0.min.js"></script> + <script src="js/vendor/bootstrap.min.js"></script> + <script src="js/main.js"></script> + + <!-- MathJax Section --> + <script type="text/x-mathjax-config"> + MathJax.Hub.Config({ + TeX: { equationNumbers: { autoNumber: "AMS" } } + }); + </script> + <script type="text/javascript" + src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script> + <script> + MathJax.Hub.Config({ + tex2jax: { + inlineMath: [ ["$", "$"], ["\\\\(","\\\\)"] ], + displayMath: [ ["$$","$$"], ["\\[", "\\]"] ], + processEscapes: true, + skipTags: ['script', 'noscript', 'style', 'textarea', 'pre'] + } + }); + </script> + </body> +</html> |