Java Numerical Library is tuned for high performance data analysis with tools for big data – providing 100% Java analytics to simplify complex code development.
Java Numerical Library 7.3 highlights
With the release of JMSL 7.3, developers have two new features for prediction and pattern recognition, SupportVectorMachine and RandomTrees, on top of new statistical classes, improvements based on customer enhancement requests, and bug fixes.
- New Support Vector Machine package
- Based on the well-known LIBSVM library, providing formulations for classification, regression, and one-class distribution estimation
- Supports the ν-SVR and Ïµ-SVR formulations for regression, the C-SVC and Ïµ-SVC formulations for classification, and the one-class formulation for goodness-of-fit
- Provides an extendable kernel class for optimization and calculation of predictions, including Linear, Polynomial, RadialBasis, and Sigmoid kernels.
- New random forest ensemble method for Decision Trees
- Based on the technique invented by Leo Breiman in 2001 to generate predictions in either classification or regression problems using a committee of decision trees, widely used in many areas including distributed, big data problems
- New RandomTrees class works with any one of the four decision tree algorithms included in the DecisionTree package
- New Statistical Classes:
- PooledCovariances: Computes the pooled variance-covariance matrix from one or more sets of observations
- RandomSamples: Generates random samples from a finite population or a specified subset of observations
- New extensions of the probability distribution class for maximum likelihood estimation: Continuous uniform and exponential
- Additional machine learning algorithms:
- Added methods to get out-of-bag predictions in BootstrapAggregation
- Added methods to estimate class probabilities and specify maximum iterations to the PredictiveModel
- Added K-Means++ algorithm to ClusterKMeans
- Added methods to improve scalability of Apriori
- Bug fixes and continued quality improvements