Benchmarking Machine Learning Tools for Scalability, Speed and Accuracy

Szilárd Pafka, PhDChief Scientist, Epoch

H2O World Conference, Mountain View Nov 2015

Disclaimer:

I am not representing my employer (Epoch) in this talk

I cannot confirm nor deny if Epoch is using or not any of the methods, tools, results etc. mentioned in this talk. The results presented in this talk should not be considered as any indication whether Epoch is using these methods, tools, results etc. or not.

I usually use other people’s code [...] it is usually not “efficient” (from time budget perspective) to write my own algorithm [...] I can find open source code for what I want to do, and my time is much better spent doing research and feature engineering -- Owen Zhanghttp://blog.kaggle.com/2015/06/22/profiling-top-kagglers-owen-zhang-currently-1-in-the-world/

Data Size for Supervised Learning

# records:<10M10M-10B>10B

Data Size for Non-Linear Supervised Learning

# records:<1M1M-100M>100M

binary classification, 10M recordsnumeric & categorical features, non-sparse

http://www.cs.cornell.edu/~alexn/papers/empirical.icml06.pdf

http://lowrank.net/nikos/pubs/empirical.pdf

http://www.cs.cornell.edu/~alexn/papers/empirical.icml06.pdf

http://lowrank.net/nikos/pubs/empirical.pdf

- R packages- Python scikit-learn- Vowpal Wabbit- H2O- xgboost- Spark MLlib

- R packages 30%- Python scikit-learn 40%- Vowpal Wabbit 8%- H2O 10%- xgboost 8%- Spark MLlib 6%

- R packages 30%- Python scikit-learn 40%- Vowpal Wabbit 8%- H2O 10%- xgboost 8%- Spark MLlib 6%- a few others

- R packages 30%- Python scikit-learn 40%- Vowpal Wabbit 8%- H2O 10%- xgboost 8%- Spark MLlib 6%- a few others

EC2

Distributed computation generally is hard, because it adds an additional layer of complexity and [network] communication overhead. The ideal case is scaling linearly with the number of nodes; that’s rarely the case. Emerging evidence shows that very often, one big machine, or even a laptop, outperforms a cluster.http://fastml.com/the-emperors-new-clothes-distributed-machine-learning/

n = 10K, 100K, 1M, 10M, 100M

Training timeRAM usageAUCCPU % by coreread data, pre-process, score test data

linear tops off(data size)

(accuracy)

linear tops off

more data & better algo

(data size)

(accuracy)

linear tops off

more data & better algo

random forest on 1% of data beats linear on all data

(data size)

(accuracy)

10x

http://datascience.la/benchmarking-random-forest-implementations/#comment-53599

we will continue to run large [...] jobs to scan petabytes of [...] data to extract interesting features, but this paper explores the interesting possibility of switching over to a multi-core, shared-memory system for efficient execution on more refined datasets [...] e.g., machine learning http://openproceedings.org/2014/conf/edbt/KumarGDL14.pdf

learn_rate = 0.1, max_depth = 6, n_trees = 300learn_rate = 0.01, max_depth = 16, n_trees = 1000

Non-Linear Supervised Learning

# records:<1M1M-100M>100M

Non-Linear Supervised Learning