Can we accelerate the path from Machine Learning R&D to operational deployment? The Internet is full of brilliant research solutions that are released under open-source, creativ...
Can we accelerate the path from Machine Learning R&D to operational deployment?
The Internet is full of brilliant research solutions that are released under open-source, creative commons, or commercial type of license agreements. Unfortunately, most solutions never make the step from R&D to operational deployment. The main reason is that these solutions are released in research code that is difficult to understand by others. In the realm of Machine Learning, the core of the solution is the trained, or untrained model, and NOT the research code in which this model was created and/or trained. To make the step from R&D to operational deployment, it is sufficient to import the model into a system that was built for operational usage.
Today’s release v.6.6.6 of OpendTect’s Machine Learning platform is a game-changer in terms of accelerating the path from R&D to operational deployment. The new version supports importing models developed in Keras (TensorFlow), Scikit Learn and PyTorch.
Our goal is to develop an extensive library with trained and untrained models that can be used to solve similar problems on datasets with only wells; only seismic; or on the combination of seismic and wells. Trained models can be re-used directly on unseen datasets by operational geoscientists. The latest generation of untrained models can be trained on proprietary data for optimal predictions in operational settings. Apart from this globally accessible library, E&P companies can use the OpendTect ML platform to create private libraries of trained and untrained models that are accessible only inside the company. We invite the E&P community to share relevant (trained or untrained) models. You decide whether you want to share your models free-of-charge, or under your own commercial terms. In the latter case, we only publish meta information about the model in our library. A user can buy your model from dGB at the price you have set. Please contact info@dgbes.com for more information.
You can also still register for our upcoming webinar about Machine Learning assisted 3D horizon tracking on 17 February at 11 am CET. Just click on this link to register: https://lnkd.in/dbtTgqcZ
We hope to see you online!
#machinelearning #geoscience #seismic #research #geophysics #PyTorch #Scikitlearn #tenserflow #python
