Large Scale Machine Learning with Apache Spark

Large Scale Learning with Apache Spark
Sandy Ryza, Data Science, Cloudera

● Data scientist at Cloudera
● Recently lead Apache Spark development at
Cloudera
● Before that, committing on Apache Hadoop
● Before that, studying combinatorial
optimization and distributed systems at
Brown
Me

Sometimes you find yourself
with lots of stuff

Two Main Problems
● Designing a system for processing huge
data in parallel
● Taking advantage of it with algorithms that
work well in parallel

System Requirements
● Scalability
● Programming model that abstracts away
distributed ugliness
● Data-scientist friendly
○ High-level operators
○ Interactive shell (REPL)
● Efficiency for iterative algorithms

CONFIDENTIAL - RESTRICTED*
MapReduce
Map Map Map Map Map Map Map Map Map Map Map Map
Reduce Reduce Reduce Reduce
Key advances by MapReduce:
•Data Locality: Automatic split computation and launch of mappers appropriately
•Fault tolerance: Write out of intermediate results and restartable mappers meant
ability to run on commodity hardware
•Linear scalability: Combination of locality + programming model that forces developers
to write generally scalable solutions to problems

Spark: Easy and Fast Big Data
• Easy to Develop
• Rich APIs in Java,
Scala, Python
• Interactive shell
• Fast to Run
• General execution
graphs
• In-memory storage
2-5× less code Up to 10× faster on disk,
100× in memory

What is Spark?
Spark is a general purpose computation framework geared towards massive
data - more flexible than MapReduce
Extra properties:
•Leverages distributed memory
•Full Directed Graph expressions for data parallel computations
•Improved developer experience
Yet retains:
Linear scalability, Fault-tolerance and Data-Locality

Spark introduces concept of RDD to take
advantage of memory
RDD = Resilient Distributed Datasets
•Defined by parallel transformations on data in stable storage

RDDs
bigfile.txt lines
val lines = sc.textFile(
“bigfile.txt”)

RDDs
bigfile.txt lines
val lines = sc.textFile
(“bigfile.txt”)
numbers
val numbers = lines.map
((x) => x.toDouble)

RDDs
bigfile.txt lines
(“bigfile.txt”)
numbers
((x) => x.toDouble)
sum
numbers.sum()

RDDs
bigfile.txt lines
(“bigfile.txt”)
numbers
Partition
Partition
Partition
Partition
Partition
Partition
HDFS
sum
Driver
((x) => x.toDouble) numbers.sum()

Shuffle
bigfile.txt lines
(“bigfile.txt”)
numbers
Partition
Partition
Partition
Partition
Partition
Partition
HDFS
sum
Driver
val sorted =
lines.sort() sorted.sum()

Persistence and Fault Tolerance
•User decides whether and how to persist
• Disk
• Memory
• Transient (recomputed on each use)
Observation:
a.Provides fault-tolerance through concept of lineage

Lineage
•Reconstruct partitions that go
down using original steps we
used to create them

RDDs
bigfile.txt lines
(“bigfile.txt”)
numbers
Partition
Partition
Partition
Partition
Partition
Partition
HDFS
sum
Driver
((x) => x.toInt) numbers.cache()
.sum()

numbers.sum()
bigfile.txt lines numbers
Partition
Partition
Partition
sum
Driver

Easy
• Multi-language support
• Interactive Shell
Python
lines = sc.textFile(...)
lines.filter(lambda s: “ERROR” in s).count()
Scala
val lines = sc.textFile(...)
lines.filter(s => s.contains(“ERROR”)).count()
Java
JavaRDD<String> lines = sc.textFile(...);
lines.filter(new Function<String, Boolean>() {
Boolean call(String s) {
return s.contains(“error”);
}
}).count();

Out of the Box Functionality
• Hadoop Integration
• Works with Hadoop Data
• Runs under YARN
• Libraries
• MLlib
• Spark Streaming
• GraphX (alpha)
• Roadmap
• Language support:
• Improved Python support
• SparkR
• Java 8
• Schema support in Spark’s APIs
• Better ML
• Sparse Data Support
• Model Evaluation Framework
• Performance Testing

So back to ML
• Hadoop Integration
• Works with Hadoop Data
• Runs under YARN
• Libraries
•MLlib
• Spark Streaming
• GraphX (alpha)
• Roadmap
• Language support:
• Improved Python support
• SparkR
• Java 8
• Schema support in Spark’s APIs
• Better ML
• Sparse Data Support
• Model Evaluation Framework
• Performance Testing

Spark MLlib
Discrete Continuous
Supervised Classification
● Logistic regression (and
regularized variants)
● Linear SVM
● Naive Bayes
● Random Decision Forests
(soon)
Regression
● Linear regression (and
regularized variants)
Unsupervised Clustering
● K-means
Dimensionality reduction, matrix
factorization
● Principal component analysis /
singular value decomposition
● Alternating least squares

Why Cluster Big Data?
● Learn the structure of your data
● Interpret new data as it relates to this
structure

Anomaly Detection
● Anomalies as data points far away from any
cluster

Train a classifier on each cluster

Using it
val data = sc.textFile("kmeans_data.txt")val parsedData =
data.map( _.split(' ').map(_.toDouble))// Cluster the
data into two classes using KMeansval numIterations =
20val numClusters = 2val clusters =
KMeans.train(parsedData, numClusters, numIterations)

K-Means
● Alternate between two steps:
o Assign each point to a cluster based on
existing centers
o Recompute cluster centers from the
points in each cluster

K-Means - very parallelizable
● Alternate between two steps:
o Assign each point to a cluster based on
existing centers
 Process each data point independently
o Recompute cluster centers from the
points in each cluster
 Average across partitions

// Find the sum and count of points mapping to each center
val totalContribs = data.mapPartitions { points =>
val k = centers.length
val dims = centers(0).vector.length
val sums = Array.fill(k)(BDV.zeros[Double](dims).asInstanceOf[BV[Double]])
val counts = Array.fill(k)(0L)
points.foreach { point =>
val (bestCenter, cost) = KMeans.findClosest(centers, point)
costAccum += cost
sums(bestCenter) += point.vector
counts(bestCenter) += 1
}
val contribs = for (j <- 0 until k) yield {
(j, (sums(j), counts(j)))
}
contribs.iterator
}.reduceByKey(mergeContribs).collectAsMap()

// Update the cluster centers and costs
var changed = false
var j = 0
while (j < k) {
val (sum, count) = totalContribs(j)
if (count != 0) {
sum /= count.toDouble
val newCenter = new BreezeVectorWithNorm(sum)
if (KMeans.fastSquaredDistance(newCenter, centers(j)) > epsilon * epsilon) {
changed = true
}
centers(j) = newCenter
}
j += 1
}
if (!changed) {
logInfo("Run " + run + " finished in " + (iteration + 1) + " iterations")
}
cost = costAccum.value

The Problem
● K-Means is very sensitive to initial set of
center points chosen.
● Best existing algorithm for choosing centers
is highly sequential.

K-Means++
● Start with random point from dataset
● Pick another one randomly, with probability
proportional to distance from the closest
already chosen
● Repeat until initial centers chosen

K-Means++
● Initial cluster has expected bound of O(log k)
of optimum cost

K-Means++
● Requires k passes over the data

K-Means||
● Do only a few (~5) passes
● Sample m points on each pass
● Oversample
● Run K-Means++ on sampled points to find
initial centers

Large Scale Machine Learning with Apache Spark

Large Scale Machine Learning with Apache Spark

More Related Content

What's hot

Similar to Large Scale Machine Learning with Apache Spark

More from Cloudera, Inc.

Recently uploaded

Large Scale Machine Learning with Apache Spark