在金融数据领域,彭博用三十多年的时间从一个向用户提供部分美国公司金融数据的公司成长到了现在覆盖全球基本上所有公司的全方面的超大型集成平台。这些金融数据都需要从不同的格式中被尽可能快速并且准确的提取出来,标准化,最后通过统一的格式反馈到市场。在本次演讲中,我们将讲述在神经网络领域最新的突破如何帮助彭博对文件进行自动化处理,并将展示其在数据提取及分析方面体现出来的更高准确度和更快处理速度。

注脚

展开查看详情

1. From Keyboards to Neural Networks 从键盘到神经网络 Qcon Beijing April 21, 2018 Biye Li Team Manager, Data Technologies Automation Xiangqian Yu Team Lead, Derivatives Data © 2018 Bloomberg Finance L.P. All rights reserved.

2. What is Bloomberg? The Bloomberg Terminal delivers a diverse array of information on a single platform to facilitate financial decision- making. 4 © 2018 Bloomberg Finance L.P. All rights reserved.

3. What is Data Technologies Automation? © 2018 Bloomberg Finance L.P. All rights reserved.

4.Challenges – Scale of Financial Information Market Types Speed To Market Accuracy AAPL FB 700 GOOG BIDU ? ? TXT ? ? Companies Problematic Files/Input Modified from https://upload.wikimedia.org/wikipedia/commons/d/dc/UnderwoodKeyboard_%28transparent%29.png https://upload.wikimedia.org/wikipedia/commons/1/18/1328102022_Document.png May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en

5. Challenges – Accuracy Really Matters Federal Reserve will maintain rate at 1.25% to 1.5% © 2018 Bloomberg Finance L.P. All rights reserved.

6. Challenges – Accuracy Really Matters Federal Reserve will maintain rate at 1.25% to 1.5% vs. Federal Reserve will raise rate to 2% © 2018 Bloomberg Finance L.P. All rights reserved.

7. Solution – Evolution Over Time patt[ern] matc[hin]g 1990s 2000s 2010 2016 2017 Data Volume Modified from https://commons.wikimedia.org/wiki/Category:Machine_learning_algorithms#/media/File:Moving_From_unknown_to_known_feature_spaces_based_on_TS-ELM_with_random_kernels_and_connections.tif https://commons.wikimedia.org/wiki/Category:Machine_learning_algorithms#/media/File:Moving_From_unknown_to_known_feature_spaces_based_on_TS-ELM_with_random_kernels_and_connections.tif https://commons.wikimedia.org/wiki/Category:Machine_learning_algorithms#/media/File:OPTICS.svg May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en © 2018 Bloomberg Finance L.P. All rights reserved.

8. Back in 2016 – Table Extraction © 2018 Bloomberg Finance L.P. All rights reserved.

9. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

10. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

11. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

12. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

13. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

14. Tables Look Different © 2018 Bloomberg Finance L.P. All rights reserved.

15. Table Detection – How Do We Do It © 2018 Bloomberg Finance L.P. All rights reserved.

16. Table Detection – How Do We Do It © 2018 Bloomberg Finance L.P. All rights reserved.

17. Table Detection – How Do We Do It © 2018 Bloomberg Finance L.P. All rights reserved.

18. Computer Vision Tasks Modified from https://commons.wikimedia.org/wiki/File:Cats_Petunia_and_Mimosa_2004.jpg © 2018 Bloomberg Finance L.P. All rights reserved. May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en

19. Computer Vision Tasks Modified from https://commons.wikimedia.org/wiki/File:Cats_Petunia_and_Mimosa_2004.jpg © 2018 Bloomberg Finance L.P. All rights reserved. May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en

20. Computer Vision Tasks Modified from https://commons.wikimedia.org/wiki/File:Cats_Petunia_and_Mimosa_2004.jpg © 2018 Bloomberg Finance L.P. All rights reserved. May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en

21. Table Detection Is Object Detection Deep learning has yielded rapid advancements in computer vision © 2018 Bloomberg Finance L.P. All rights reserved.

22. CNN Modified from https://commons.wikimedia.org/wiki/File:Typical_cnn.png © 2018 Bloomberg Finance L.P. All rights reserved. May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en

23. ResNet-152 Building Block Repeat this 50 times © 2018 Bloomberg Finance L.P. All rights reserved.

24. Evolution of Depth AlexNet 8 Layers ILSVRC 2012 VGG 19 Layers ILSVRC 2014 ResNet 152 Layers ILSVRC 2015 © 2018 Bloomberg Finance L.P. All rights reserved.

25. Faster RCNN • RCNN bounding-box region proposals classification regression • Fast RCNN classification region proposals bounding-box regression • Faster RCNN region proposals classification bounding-box regression © 2018 Bloomberg Finance L.P. All rights reserved.

26. Faster RCNN Region Proposal Network Conv Layers Feature Maps Classifier Faster RCNN https://github.com/rbgirshick/py-faster-rcnn © 2018 Bloomberg Finance L.P. All rights reserved.

27. Faster RCNN © 2018 Bloomberg Finance L.P. All rights reserved.

28. Visualizing Neural Network © 2018 Bloomberg Finance L.P. All rights reserved.

29. Object Detection Good Enough? Modified from https://commons.wikimedia.org/wiki/File:Cats_Petunia_and_Mimosa_2004.jpg © 2018 Bloomberg Finance L.P. All rights reserved. May be re-distributed in accordance with the terms of the CC-SA 4.0 license https://creativecommons.org/licenses/by-sa/4.0/deed.en