I placed 1360 in a deep learning competition and here’s why that is a win in my book.
I spend too much time on social media, but this time I was lucky. I checked my feed on LinkedIn quickly and saw an announcement from TGS a seismic contractor. They had teamed up with Kaggle, a machine learning competition site that warped into a learning resource and community. They host a competition to identify salt in seismic images and put out $100,000 US in prize money.
So I wrote a solution, submitted my prediction and placed 5th. That changed fast. After sleeping a bit, I dropped to position 600. Today I would be below 1300 if I had not improved upon my first solution.
Why does this feel like a win?
Well, I decided to write an open solution with an introduction to salt in seismic data for those coming from the machine learning side and implement a complete deep learning workflow for those coming from a geoscience background.
What makes it feel even more like a win?
The reactions showed that many believed I was, in fact, part of either Kaggle or TGS, due to speed and quality. Several forks build upon my initial solution and improve it significantly. There are 400 upvotes on the kernel, which is fantastic to me. And 25,000 views were counted on my kernel, which is mind-boggling. Also, 700 copies of my solution have been made. This is probably the most impactful thing I have done in my life.
Check it out over here: https://www.kaggle.com/jesperdramsch/intro-to-seismic-salt-and-how-to-geophysics
Latest posts by Jesper Dramsch (see all)
- Research Talk — Deep Learning for 4D Pressure Saturation Inversion [Youtube] - 2019-04-18
- Geysers in Slow Motion - 2019-02-04
- Keynote Bonanza and No Coffee – The EAGE / PESGB ML Workshop - 2018-12-17