Lessons from 2MM machine learning models
Download as PPTX, PDF7 likes61,212 views
The document outlines various data science competitions hosted on Kaggle, including challenges with substantial cash prizes, like optimizing flight routes and developing automated scoring algorithms. It highlights the achievements of prominent participants, such as a PhD student who surpassed established algorithms in the mapping dark matter competition. Additionally, it notes trends in competition participation and emerging methodologies in machine learning, specifically in structured data and computer vision.
Lessons from 2MM machine learning models
- 1. Kaggle The home of data science
- 2. GE Flight Quest 2 Optimize flight routes based on weather & traffic $250,000 122 teams Hewlett Foundation: Automated Essay Scoring Develop an automated scoring algorithm for student-written essays $100,000 155 teams Allstate Purchase Prediction Challenge Develop an automated scoring algorithm for student-written essays $50,000 1,570 teams Merck Molecular Activity Challenge Help develop safe and effective medicines by predicting molecular activity $40,000 236 teams Higgs Boson Machine Learning Challenge Use the ATLAS experiment to identify the Higgs boson $13,000 1,302 teams
- 3. Age Income Default 58 $95,824 True 73 $20,708 False 59 $82,152 False 66 $25,334 True Age Income Default 73 $53,445 61 $36,679 47 $90,422 44 $79,040 Training Data Test Data The Kaggle Approach
- 5. Mapping Dark Matter Competition Progress Accuracy (lower is better) Week 1 Week 3 Week 5 Week 7 End .0150 .0170 Martin O’Leary PhD student in Glaciology, Cambridge U
- 6. “In less than a week, Martin O’Leary, a PhD student in glaciology, outperformed the state-of-the-art algorithms” “The world’s brightest physicists have been working for decades on solving one of the great unifying problems of our universe”
- 7. Mapping Dark Matter Competition Progress Accuracy (lower is better) Week 1 Week 3 Week 5 Week 7 End .0150 .0170 Martin O’Leary PhD student in Glaciology, Cambridge U Marius Cobzarenco Grad student in computer vision, UC London Ali Haissaine & Eu Jin Loc Signature Verification, Qatar U & Grad Student @ Deloitte Other deepZot (David Kirkby & Daniel Margala) Particle Physicist & Cosmologist
- 8. We’ve worked with many of the world’s largest companies Healthcare & Pharma Consumer Internet Finance IndustrialConsumer Marketing Oil & Gas $50b+ Beverage Co. Global Bank Top Credit Card Issuer Top 5 E&P Top 20 E&P
- 10. That submit over 100K machine learning models per month 0 20,000 40,000 60,000 80,000 100,000 120,000 140,000 160,000 May-10 May-11 May-12 May-13 May-14 May-15 Monthly Submissions to Kaggle Competitions
- 11. There’s a cookbook for winning competitions on structured data. It starts with exploring the data.
- 12. 2. Create and select features
- 14. A second cookbook is emerging on computer vision and speech problems. It involves using convolutional neural networks.
- 15. The vast majority of time is spent training algorithms when CNNs are applied.
- 16. There are the problems that land in the middle…
Editor's Notes
- #3: People currently come to Kaggle
- #5: We score their solutions in real time.
- #9: People don’t come to us with churn or cross sell, but they typically come to us with their hardest problems, and I’ll talk more about this soon. It’s for these reasons that we continue to invest in the competition platform. It’s a very efficient operation. It’s currently running with a headcount of 4. We believe 6 is the right long term number of people to invest in competitions. We decided to focus on Oil & Gas because after working with ~25 Fortune 500s and 12 industries, we believe it’s the biggest opportunity for machine learning and most ripe for disruption. Specifically because: Greatest value add: Huge gap between what they’re doing and what’s possible Shale is disruptive: the industry is looking for new ideas making it a good environment to be selling into.
- #10: We score their solutions in real time.
- #11: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #12: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #13: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #14: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #15: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #16: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media
- #17: Kaggle Competitions – breakeven business Access to most advanced and proven techniques Recruiting the very best of a scarce resource C-level access from leadership positioning in media