Aristides Zenonos, Ph.D.

I would like to share the paper that was just published in this month's issue of First Break, of which I am a co-author. The paper shows several examples illustrating the value of the PS wave in elastic inversions. Thanks to dr Ali Tura for inviting me to contribute to this work! https://lnkd.in/dzw5ftV9

72

5 Comments

Transforming unstructured data generated by LLMs into structured formats is an important task for many applications. Using Pydantic and LangChain, developers can seamlessly convert raw text into structured data like JSON, enabling easier downstream processing and integration.

14

Great talk! A stream processor does 3 jobs - coordinates input streams -> transform data 0 -> convert to useful formats - with necessity of historical and real-time/streaming modes. Existing open-source python packages for streaming data: - bytewax: written in Rust, pre-written logic for control flow& statictics (ma, weighted ma, etc), supports multiple data formats connectors - Apache Flink: complex but most flexible - Apache Beam: parallelization - .... CSP: python API with C++ runtime, no infra constraints 1. seamless historical/real-time transition 2. No parallelization (yet)

1

If you’re going to EGU come along to see Prof Jieping Ye, a specialist in machine learning, discuss some of the opportunities and challenges of large language models for geoscience, including DDE’s GeoGPT an LLM which is currently in development. A large language model (LLM) is a type of artificial intelligence (AI) program that can recognize and generate text. LLMs are trained on huge amounts of data — hence the name "large." LLMs are built on machine learning - a type of neural network called a transformer model. In this session we introduce DDE’s ‘GeoGPT’ LLM. GeoGPT is still in development and not on general release, but its performance as tested by many scientists is excellent. It is the first geoscience LLM, and an LLM built for geologists by geologists. Some other notable LLMs are OpenAI's GPT series of models (e.g., GPT-3.5 and GPT-4, used in ChatGPT and Microsoft Copilot), and Google's PaLM and Gemini. Studies have shown that LLMs like ChatGPT and GeoGPT may create issues with the attribution of credit and rights for text generation, as well as questions over responsibility and liability in relation to any harmful consequences that might result from the use of LLM generated material. For example, who takes responsibility for incorrect or misleading material produced by an LLM – the creator of the LLM - or the user who passes misleading information on to a third party? This is just one of the ethics and philosophical  issues that LLMs have created, and our session will foster an open discussion, as well as a chance to hear more about GeoGPT.

9

🚀 Demystifying Multiclass Classification: OvA vs. OvO Strategies! 🤖 When tackling real-world classification problems, we often move beyond simple binary outcomes (e.g., spam vs. not spam) to multiclass classification tasks, where there are multiple possible labels (e.g., handwritten digits 0-9). But did you know that there are specific strategies to adapt binary classifiers for these tasks? Enter One-versus-All (OvA) and One-versus-One (OvO)! 🔹 One-versus-All (OvA) This approach trains one binary classifier per class. For example, if you're classifying images of digits, you'd have 10 classifiers: one for each digit (0 vs. all, 1 vs. all, etc.). The input is passed through each classifier, and the class with the highest decision score is chosen. Pros: Simpler setup with fewer models. Best For: Algorithms that scale well with the size of the training set. 🔹 One-versus-One (OvO) In OvO, a binary classifier is trained for each pair of classes, resulting in N×(N−1)/2N \times (N - 1) / 2N×(N−1)/2 classifiers (e.g., for 10 classes, that's 45 models!). During prediction, the input is evaluated by all classifiers, and the class with the most "wins" is selected. Pros: Each model is trained on smaller subsets, which is ideal for algorithms like SVMs that don’t scale well with large datasets. Best For: When training many smaller models is preferable to training fewer, larger ones. Understanding these strategies not only helps optimize your machine learning workflow but also deepens your appreciation for how algorithms scale and perform across various tasks. Next time you face a multiclass problem, remember that your choice between OvA and OvO can make all the difference! 💡 Which strategy do you prefer for multiclass problems? Share your thoughts below! 👇 #MachineLearning #DataScience #MulticlassClassification #OvA #OvO #AI

1

I am pleased to share our recent paper on quantifying well leakage with Matteo Loizzo. This is the first of a two part paper, focusing on modelling the initial state of cement in wells. In part two, we will elaborate on the calculation of well leaks based on the modelling results and present a case study for a CCS well. #wellleakage #wellintegrity

63

3 Comments

The global Large Language Model (LLM) Market is experiencing significant growth, with projections indicating an increase from USD 6.4 billion in 2024 to USD 36.1 billion by 2030, driven by a 33.2% CAGR. This expansion is fueled by rising demand for advanced natural language processing and the advancements in cloud computing, which enhance the deployment and training of these models across various industries.

2

Pack just released a new book on the practical application of graph theory. It provides valuable explanations and practical examples of how graph-based algorithms implemented using Python solve industry problems, inspiring and motivating you to apply these techniques in your own work. It is an excellent book to expand your machine learning programming skills around network science. If you are interested in this book, please follow this link: https://lnkd.in/g2mD_CFY

30

5 Comments

Convolutional neural networks can help geophysicists identify hydrocarbon traps within salt bodies and their surrounding sediments. Learn how DS365.ai helps make it possible: https://lnkd.in/dAmee_Rb #convolutionalneuralnetworks #geophysics Halliburton

1

Sharing a technical report titled "LoRA Land: 310 Fine-tuned LLMs that Rival GPT-4, A Technical Report" which is recently published on arXiv. Check out the report to gain insights on how LoRA fine-tuning can help open-source LLM to outperformed GPT-4. Read the full report here: https://lnkd.in/gV-uN-Gp #artificialintelligence #LLM #LORA

5

Using Vortexa's Python SDK https://lnkd.in/ehsdqsFn you can get lightning-fast access to complete global waterborne #oilandgas flows across time, from the level of berths and terminals up to countries and shipping regions. You can then integrate this #data as powerful features to your internal #data and #machine_learning pipelines to better understand various market trends and patterns and #forecast key market indicators. Book a demo to find out more: https://bit.ly/3Rr8Mb1 #Vortexa #SDK #Python #datascience

51

3 Comments

🔐 The potential of utilizing feature importance in clustering often goes unrecognized. However, with careful observation, it can greatly enhance our understanding of the underlying knowledge and patterns within our dataset. 📕 Explore its application on geoscience domain in this article: https://lnkd.in/gtabRGk2

25

Exciting Project Alert: From KNN to VGG16 - A Modern Approach to Image Expression Classification Using Neural Networks and Transfer Learning I’m thrilled to share my latest project, where I explored the fascinating world of image expression classification, specifically focusing on distinguishing between happy and sad faces, leveraging the power of Convolutional Neural Networks (CNNs) and Transfer Learning. Project Overview: In this project, I classified facial expressions into happy and sad categories using various models. Starting from traditional methods like K-Nearest Neighbors (KNN) and Support Vector Machines (SVM), I advanced to Random Forest and finally applied VGG16 through Transfer Learning. The convolutional layers in CNN were pivotal in extracting features such as smiles, frowns, and other subtle facial cues that are crucial for accurately classifying expressions. Transfer Learning allowed me to harness the power of the pre-trained VGG16 model, significantly improving the classification accuracy by leveraging extensive prior training on large datasets. 🔧 Key Highlights: Feature Extraction with CNNs: Automatically learned essential features for expression recognition, such as facial landmarks and texture. Transfer Learning: Applied VGG16, a deep learning model with extensive training on diverse image datasets, to improve performance in classifying facial expressions.Performance Optimization: Techniques like GridSearch and cross-validation were employed to fine-tune the models, ensuring the best possible accuracy. #MachineLearning #DeepLearning #CNN #TransferLearning

6

1 Comment

📍Kolmogorov-Arnold Networks - Chapter 1 “100x more accurate and parameter efficient than MLPs at modelling ODEs” - Math and how it compares to a Multi Layer Perceptron - Architecture and KAT (approximation theorem) - Challenges In the next post(s) - applications, use cases and limitations

44

3 Comments

Here we go 🎉🎯 The First phase for the #CCS #Ravenna project (JV #Snam #Eni ) is currently operational, injecting #CO2 with one well in the subsurface of a depleted gas field, offshore Ravenna, Northern Adriatic 🇮🇹. Hopefully, this is just the starting point, leading to a near future large scale industrial project at the Megatonne scale. https://lnkd.in/d9WzDr-3 https://lnkd.in/dWiDU73u

150

8 Comments

I will walk through a tutorials of articles about how you can apply your python basic knowledge into a practical and very nice examples, ISA. In this article, i talk about an example of capturing uncertainty in well production rates using Python, OpenServer, and Prosper. I explore the use of Monte Carlo simulations to estimate production rates, providing a probabilistic distribution that offers a more comprehensive risk assessment than traditional methods. This examaple show how predicting well production rates accurately is paramount for optimizing well test design and making informed investment decisions Check out the full blog post for a detailed guide and let me know your thoughts and recommendations! https://lnkd.in/dDGkRV5h let me know if you want the full notebook #OilAndGas #Python #MonteCarloSimulation #Prosper #OpenServer #ReservoirEngineering #DataScience

43

8 Comments

I just presented the paper: The Integration of Shallow to Ultra-Deep LWD Data, the Key to Geosteering and Improved Reservoir Understanding. At #SPWLA2024, which can be found at https://lnkd.in/dWgfk-AD Thanks to the coauthors

133

14 Comments

In this session, we are going to showcase various applications of the results of SEAM projects throughout multiple years. If you want to know more, attend this half day session during #IMAGE24 by registering in the link below. #SEG#geophysics

25