[Paper Review] Quantum Lithology Interpretation

Interpreting lithology is critical for achieving a greater understanding of subsurface properties. Manual interpretation is a time-consuming activity that might lead to inconsistency, where the most used conventional methods are cross-plotting and statistical analysis-based techniques. Some papers propose using machine learning (SVM, NNs) and deep learning (CNN, RNN) applications of various types.

However, deep learning-based (CNN) models require large training datasets for generalization, and obtaining a sufficient amount of training data with labels is time-consuming and expensive.

The research paper, "Naihao Liu et al., 2021, Quantum-Enhanced Deep Learning-Based Lithology Interpretation From Well Logs," proposes to improve machine-learning and deep-learning algorithms by integrating parameterized quantum circuits. Quantum computing makes use of superposition and entanglement that benefit deep-learning models with greater generalization potential.

The authors tested the quantum technique on field data and compared the lithology interpretation results to deep-learning-based methods (CNN, LSTM). It was discovered that with fewer training examples, the quantum-based approach performed marginally better than others. Moreover, the authors emphasize that the suggested technique is distinctly beneficial in assessing thin lithology layers.

Link for the research paper
Link for the LinkedIn post