© 2018 GridGain Systems, Inc. Distributed Machine Learning with Zero ETL Yury Babak Head of development, GridGain
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
© 2018 GridGain Systems, Inc.
Distributed Machine Learning with Zero ETL
Yury Babak
Head of development, GridGain
© 2018 GridGain Systems, Inc.
Long ETL
© 2018 GridGain Systems, Inc.
Long ETL
- Х%
- Х%
© 2018 GridGain Systems, Inc.
Distributed Training
© 2018 GridGain Systems, Inc.
Node Crash
© 2018 GridGain Systems, Inc.
Apache Ignite
© 2018 GridGain Systems, Inc.
Apache Ignite: Replicated Caches
Server Node 1 Server Node 2
Server Node 3 Server Node 4
Client
© 2018 GridGain Systems, Inc.
Map Reduce
© 2018 GridGain Systems, Inc.
Iterative Optimization Algorithm
© 2018 GridGain Systems, Inc.
Partition Based Data Set
© 2018 GridGain Systems, Inc.
Restoration of partitions after a failure
© 2018 GridGain Systems, Inc.
Recovering calculations after failure
© 2018 GridGain Systems, Inc.
OLS sample
Loss function
Gradient of loss function
Node 2Node 1Node M
© 2018 GridGain Systems, Inc.
Sample 2 LSQR
© 2018 GridGain Systems, Inc.
Limitations of Applicability
Iteration time
Number of Iterations
SGDBS 1 000
BS 10
Time to training
© 2018 GridGain Systems, Inc.
https://ignite.apache.org
https://apacheignite.readme.io/docs
https://github.com/apache/ignite
Want to learn more?
Related Documents
