[SPARK-9471] Multilayer perceptron classifier

• Type: New Feature
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.4.0
• Fix Version/s: 1.5.0
• Component/s: ML
• Target Version/s: 1.5.0
• Assignee: Alexander Ulanov
• Reporter: Alexander Ulanov
• Shepherd: Xiangrui Meng
• Created: 30/Jul/15 10:34
• Updated: 12/Jan/16 02:28
• Resolved: 31/Jul/15 18:25

Description

Implement Multilayer Perceptron for Spark ML. Requirements:
1) ML pipelines interface
2) Extensible internal interface for further development of artificial neural networks for ML
3) Efficient and scalable: use vectors and BLAS

Issue Links

• incorporates: SPARK-4752 Classifier based on artificial neural network
• relates to: SPARK-10583 Correctness test for Multilayer Perceptron using Weka Reference
• relates to: SPARK-10478 Improve spark.ml.ann implementations for MLP
• relates to: SPARK-11262 Unit test for gradient, loss layers, memory management for multilayer perceptron
• relates to: SPARK-12664 Expose probability, rawPrediction in MultilayerPerceptronClassificationModel
• relates to: SPARK-5575 Artificial neural networks for MLlib deep learning
• relates to: SPARK-9897 User Guide for Multilayer Perceptron Classifier
• links to: [Github] Pull Request #7621 (avulanov)