GNGTS 2021 - Atti del 39° Convegno Nazionale

503 GNGTS 2021 S essione 3.3 Conclusion A great strength of this method is its sensibly shorter prediction time as compared to tradi- tional approaches. LSTM is able to work on any record length and we analyzed also the change in prediction time depending on such parameter, resulting, as expected, in a linear relationship between record length and time of prediction. The prediction time is 0.0025 s for a single trace with 1024 time samples on a machine with a 2 core Intel(R) Xeon(R) CPU, 2.20GHz and a Nvidia Tesla T4 GPU. The test performed confirmed the high versatility of the method which is totally automated and gives prediction associated to a probability value, which in turn automatically quantifies the reliability of each obtained result. The 1-D approach limits the overall subjectivity of the proce- dure, without degrading the quality of the extracted horizons on 2-D and even on 3-D datasets. Further researched will be addressed to horizon patching and automated phase assessment. Acknowledgements This research was supported by PNRA projects IPECA (PNRA18_00186) and CRIOVEG (PNRA18_00288) and by the project “Dipartimento di Eccellenza” of the Department of Mathe- matics and Geosciences of the University of Trieste. We gratefully acknowledge the support of Halliburton Landmark through the University Grants Program (UGP). References Chetlur, Sharan, Cliff Woolley, Philippe Vandermersch, Jonathan Cohen, John Tran, Bryan Catanzaro, and Evan Shelhamer. “CuDNN: Efficient Primitives for Deep Learning.” ArXiv:1410.0759 [Cs], December 17, 2014. http://arxiv.org/abs/1410.0759. Colucci, R. R., E. Forte, C. Boccali, M. Dossi, L. Lanza, M. Pipan, and M. Guglielmin. “Evaluation of Internal Structure, Volume and Mass of Glacial Bodies by Integrated LiDAR and Ground Penetrating Radar Sur- veys: The Case Study of Canin Eastern Glacieret (Julian Alps, Italy).” Surveys in Geophysics 36, no. 2 (March 2015): 231–52. https://doi.org/10.1007/s10712-014-9311-1. Forte, E., Dossi, M., Pipan, M., Del Ben, A. “Automated phase attribute-based picking applied to reflection seismics”. Geophysics, 81, no. 2, V141-V150. https://doi.org/10.1190/geo2015-0333.1 Geletti, R., F. Zgur, A. Del Ben, F. Buriola, S. Fais, M. Fedi, E. Forte, et al. “The Messinian Salinity Crisis: New Seismic Evidence in the West-Sardinian Margin and Eastern Sardo-Provençal Basin (West Mediterra- nean Sea).” Marine Geology 351 (May 2014): 76–90. https://doi.org/10.1016/j.margeo.2014.03.019. Guo, R, Zhang, J.J., Liu, D., Zhang, Y.B. and Zhang, D.W “Application of Bi-directional Long Short-Term Me- mory Recurrent Neural Network for Seismic Impedance Inversion.” European Association of Geoscien- tists & Engineers 1 (2019): 1–5. https://doi.org/10.3997/2214-4609.201901386. Hochreiter, Sepp, and Jürgen Schmidhuber. “Long Short-Term Memory.” Neural Computation 9, no. 8 (No- vember 1, 1997): 1735–80. https://doi.org/10.1162/neco.1997.9.8.1735. Hughes, Tyler W., Ian A. D. Williamson, Momchil Minkov, and Shanhui Fan. “Wave Physics as an Analog Recurrent Neural Network.” Science Advances 5, no. 12 (December 2019): eaay6946. https://doi. org/10.1126/sciadv.aay6946. Kavzoglu, Taskin. “Increasing the Accuracy of Neural Network Classification Using Refined Training Data.” Environmental Modelling & Software 24, no. 7 (July 2009): 850–58. https://doi.org/10.1016/j. envsoft.2008.11.012. Kingma, Diederik P., and Jimmy Ba. “Adam: A Method for Stochastic Optimization.” ArXiv:1412.6980 [Cs], January 29, 2017. http://arxiv.org/abs/1412.6980. Mannor, Shie, Dori Peleg, and Reuven Rubinstein. “The Cross Entropy Method for Classification.” In Procee- dings of the 22nd International Conference on Machine Learning - ICML ’05, 561–68. Bonn, Germany: ACM Press, 2005. https://doi.org/10.1145/1102351.1102422. Martin, Gary. “The marmousi2 model, elastic synthetic data, and an analysis of imaging and AVO in a struc- turally complex environment,” n.d., 226.