Apache Mahout

Apache Mahout was created to help Java developers write scaling machine learning algorithm.

Mahout is meant to handle huge data loads, along with business-level code, all under an open source license, but at commercial-grade results.

Currently the Mahout process is as follows: mine data to acquire it, cluster it to organize the data and classify it to learn it.

What is new in this release:

• This release fixes a major memory usage bug in co-occurrence analysis used by the driver spark-itemsimilarity MAHOUT-1707. This will now require far less memory in the executor.
• Support Spark 1.2.2 or less - due to a bug in Spark 1.2+ in the JavaSerializer (SPARK-6069) we removed the use of Guava from any code executed in Spark Executors. To do this we created a Scala Collections based BiMap so any example code showing how to use the old Guava collections is obsolete.
• Some minor fixes to Mahout-Samsara QR Decomposition and matrix ops.
• Trim down packages size to < 200MB - MAHOUT-1704.
• Minor testing indicates binary compatibility with Spark 1.3 except for the Mahout Shell, which does not run.

What is new in version 0.10.1:

• This release fixes a major memory usage bug in co-occurrence analysis used by the driver spark-itemsimilarity MAHOUT-1707. This will now require far less memory in the executor.
• Support Spark 1.2.2 or less - due to a bug in Spark 1.2+ in the JavaSerializer (SPARK-6069) we removed the use of Guava from any code executed in Spark Executors. To do this we created a Scala Collections based BiMap so any example code showing how to use the old Guava collections is obsolete.
• Some minor fixes to Mahout-Samsara QR Decomposition and matrix ops.
• Trim down packages size to < 200MB - MAHOUT-1704.
• Minor testing indicates binary compatibility with Spark 1.3 except for the Mahout Shell, which does not run.

What is new in version 0.10.0:

• Scala DSL Bindings for Mahout Math Linear Algebra.
• Recommenders as a Search.
• Support for easy functional Matrix views and derivatives
• JSON output format for ClusterDumper
• Enable randomised testing for all Mahout modules using Carrot RandomizedRunner.
• Online Algorithm for computing accurate Quantiles using 1-dimensional Clustering.
• Upgrade Mahout to Lucene 4.6.1

What is new in version 0.8:

• Support for biased item-based recommender
• SGD matrix factorization for rating prediction with user and item biases
• Support for SVD++
• Support for converting one or more Lucene storage indexes to SequenceFiles as well as an upgrade of the supported Lucene version to Lucene 4.3.1.
• New streaming k-means implementation that offers on-line (and fast) clustering
• Make conversion to SequenceFiles Map-Reduce, 'seqdirectory' can now be run as a MapReduce job.
• Added an option to MinHashDriver that specifies the dimension of vector to hash (indexes or values).
• Matrix Concat utility, presently only concatenates two matrices.
• Upgraded to CommonsLang3
• Speedup the Mahout build by making tests run in parallel.