Yubetsu’s Post

What do you know about Knowledge Graphs? Check out our latest blog post to learn about this promising tool and it's role in the future of AI. #AI #knowledgegraphs #naturallanguageprocessing #nlp

New Blog Post 🤩 Do you know how to use a Knowledge Graph? We talk about their history and future AI significance here... #AI #AIBI #knowledgegraphs #naturallanguageprocessing #nlp #qualitativedata #thematicanalysis #literaturereview #rag

Balancing Tradition and Innovation: Node Features in GNNs This blog focuses on feature design for graph-based machine learning, emphasizing the importance of capturing the relational structure of… Continue reading on Medium »

New Blog Post 🚀 Do you know if a Knowledge Graph is important or not? Knowledge Graphs: The Backbone of Modern Information Systems In the era of big data and artificial intelligence, the way we organize, interconnect, and derive meaning from information has become increasingly crucial. Enter knowledge graphs: a powerful tool that has revolutionized how we represent and utilize complex, interconnected data. This article delves into the concept of knowledge graphs, their historical context, current applications, and future potential, with a particular focus on their relevance to academic research and information systems. https://lnkd.in/g8PKYKDq

Building a simple RAG is easy: many existing tools - both open-source and commercial - allow you to get started. However, reaching a production-ready state is much tougher: there are many components that require tweaking and adapting, and it's very hard to evaluate the final output automatically. This RAG talk, which was given on April 24th at the Vienna Deep Learning Meetup, overviews some of the potential tweaks, focusing especially on the data retrieval part. These RAG state-of-the-art slides are based on several sources that are highly recommended to read if you are about to build a RAG on your own. Check them out on the final slide

Eryk Lewinson brings us 𝗘𝘅𝗽𝗲𝗿𝗶𝗺𝗲𝗻𝘁 𝗧𝗿𝗮𝗰𝗸𝗶𝗻𝗴 & 𝗛𝘆𝗽𝗲𝗿𝗽𝗮𝗿𝗮𝗺𝗲𝘁𝗲𝗿 𝗧𝘂𝗻𝗶𝗻𝗴: 𝗞𝗲𝗲𝗽𝗶𝗻𝗴 𝗬𝗼𝘂𝗿 𝗧𝗿𝗶𝗮𝗹𝘀 𝗢𝗿𝗴𝗮𝗻𝗶𝘇𝗲𝗱 𝘄𝗶𝘁𝗵 𝗗𝗩𝗖 ️ As data scientists or machine learning engineers, keeping track of experiments, especially during 𝗛𝘆𝗽𝗲𝗿𝗽𝗮𝗿𝗮𝗺𝗲𝘁𝗲𝗿 𝗧𝘂𝗻𝗶𝗻𝗴 (𝗛𝗣𝗧), can be challenging. In this 𝗮𝗿𝘁𝗶𝗰𝗹𝗲, Eryk explores how DVC can be leveraged to ensure reproducibility and organization throughout the entire ML experiment lifecycle, including HPT on Towards Data Science. He explored: 🔹𝗠𝗮𝗻𝘂𝗮𝗹 𝗛𝗣𝗧: Versioning 𝑝𝘢𝑟𝘢𝑚𝘴.𝘺𝑎𝘮𝑙 for control and tracking. 🔹𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 𝗔𝗽𝗽𝗿𝗼𝗮𝗰𝗵𝗲𝘀: Exhaustive/randomized grid search with  𝑑𝘷𝑐 𝑟𝘶𝑛 𝑒𝘹𝑝. 🔹𝗕𝗮𝘆𝗲𝘀𝗶𝗮𝗻 𝗚𝗿𝗶𝗱 𝗦𝗲𝗮𝗿𝗰𝗵: Utilizing Optuna with 𝐷𝘝𝐶𝘓𝑖𝘷𝑒𝘊𝑎𝘭𝑙𝘣𝑎𝘤𝑘 for seamless tracking. By combining these techniques with DVC's core functionalities, we can ensure all experiments, including HPT trials, are fully reproducible. This article is a valuable resource for anyone looking to streamline their HPT workflows and maintain control over their ML experiments. Follow DVC.ai

How can a Nobel Laureate physicist’s approach to problem-solving help data scientists reveal the business impact of ML models? In my latest Medium article, I explore how Enrico Fermi’s estimation techniques—famous for breaking down complex problems into manageable insights—can empower data scientists and product managers to make quick, effective business forecasts with machine learning. This piece takes a step away from my previous deep technical dives and offers a fresh perspective on integrating scientific reasoning into business decisions. I hope you find it insightful! https://lnkd.in/gqgKDMEP

🚀 Exciting News for Time Series Classification Enthusiasts! If you're working on Time Series Classification, especially with deep learning models, and you're familiar with our efforts to reduce model complexity with the LITE classifier [1], we have some great news! 🌟 Our extended work on this topic has been accepted in the International Journal of Data Science and Analytics 📰, in their special issue on Learning from Temporal Data! 🎉 📄 In our paper, "Look Into The LITE In Deep Learning For Time Series Classification", we introduce a multivariate specific version of LITE, called LITEMV, and demonstrate its application in human rehabilitation assessment. 💡 LITE and LITEMV, as well as their ensemble versions, are both implemented in aeon , you can checkout how to use them on the documentation: https://lnkd.in/exYjr3tE preprint on arxiv: https://lnkd.in/eTqGg2Bb source code: https://lnkd.in/eyTTxYHC Thanks to the co-authors on this work: Maxime Devanne , stefano Berretti , Jonathan Weber and Germain Forestier [1] Ali Ismail-Fawaz, Maxime Devanne, Stefano Berretti, Jonathan Weber, Germain Forestier. "Lite: Light inception with boosting techniques for time series classification" 2023 IEEE 10th International Conference on Data Science and Advanced Analytics (DSAA) #timeseries #timeseriesclassification #deeplearning #timeseriesmachinelearning #datamining #timeseriesanalysis

When I started reading Chapter 2 of Hands-On Machine Learning, I had one simple goal: boil it down to 30 meaningful questions that capture the essence of the entire chapter. I’m trying to do this with every chapter I read—because great questions lead to deeper thinking, better understanding, and more engaging conversations. Here’s Batch 2, which focuses on the dataset and its role in ML projects: 1️⃣ What are the key features of the California Housing Prices dataset, and why are they relevant for the prediction task? 2️⃣ Why is the California Housing Prices dataset considered suitable for educational purposes? 3️⃣ What challenges might arise when transitioning from artificial datasets to real-world data? It’s fascinating how much rides on the data we use. Get it right, and everything flows. Get it wrong, and the model doesn’t stand a chance. Next up: data pipelines, where we transform raw data into actionable insights. Stay tuned!

We are excited to announce, in AI Flow Solutions, that we have open-sourced the library, MADS. It is available on our GitHub repository: https://lnkd.in/dapmMydw. The library is also available on PyPi, and you can test it out using `pip install pymads`. For detailed usage instructions, check out the README in the repository. Additionally, this repository includes the code implementation of our recently published paper, "Towards Universal Applied Machine Learning: A Multi-Agent Framework for Systematic Pipeline Executions" https://lnkd.in/dckewiye. To explore the experimental results presented in the paper, please check the "experimental-results-paper" branch in the repository, which contains the exact experiments and results discussed.