It’s the time of the year. The large applied geoscientists conference is coming up, the EAGE Annual Meeting. This year the EAGE is coming to London. I’ll be there as well and I’ll have a couple of speaking engagements, I’d like to invite you along to.
Hackathon and Sprint
I love the newly created ritual of creating an off-brand hackathon accompanying the EAGE annual event. This year, Agile* stepped down as an organizer, but still stepped up as a sponsor for the London hack.
I am planning to work on two general projects. On June first, there is a global TensorFlow Docs Sprint that I want to spend some time on.
I’ll present work from my collaboration with ETLP Team at Heriot Watt University on including physics in the neural network architecture. I think Lukas Mosser, Olivier Dubrule, Mark Thompson, and Duncan Irving have put a lot of thought in this event. I’m looking forward to an insightful and inspiring event. Just take a look at the (abridged) description:
📝 The workshop will discuss recently developed applications of ML, and the challenges and opportunities associated with the development of these applications in the petroleum industry. The first half of the workshop will include:
- Speed Posters
- Where Artificial Intelligence and DL are going in the industry (digital leaders / major operators)
- Technical presentations (petroleum companies / academia)
The second half of the workshop will be dedicated to:
- Recent industry initiatives on data availability, open communities, cloud computing and training
- Perspectives from outside the industry
- Structured, themed and interactive exchanges with the audience.
My talk will be focusing on:
Including Physics in Deep Learning
An example from 4D seismic pressure saturation inversion
Geoscience data often have to rely on strong priors in the face of uncertainty. Additionally, we often try to detect or model anomalous sparse data that can appear as an outlier in machine learning models. These are classic examples of imbalanced learning. Approaching these problems can benefit from including prior information from physics models or transforming data to a beneficial domain. We show an example of including physical information in the architecture of a neural network as prior information. We go on to present noise injection at training time to successfully transfer the network from synthetic data to field data.
Tuesday and Wednesday Career Advice
Someone thought it’s a good idea for me to be on a stage and give career advice. I find this relatively funny, as I’m mainly bumbling my way around science, while being pleasant to work with. If you’d like to hear me speak anyways, come by the EAGE Career Advice Center.
On Tuesday, June 4th at 10:00, I hope to have a conversation on
On Wednesday, June 5th around lunchtime at 12:30, we’ll chat around
Making sense of AI for a career in a changing industry
I’ll see you at ExCel in London!
Latest posts by Jesper Dramsch (see all)
- 2020 Fast Approaching! – Friday Faves - 2020-01-03
- Machine Learning for Science – A Youtube Series - 2019-12-30
- Something for a Long Trip or Unwind during the Holidays – Friday Faves - 2019-12-20