Paul Tiwald

[New on our blog] How to Automate ML Experiment Management With CI/CD Authors: Kilian Kluge and Dhruvil Karani Reading time: 8 min — (link to the full article in the comments) #ml #machinelearning #mlops

2

1 Kommentar

Looking to optimize your RAG retrieval system but short on time? We've got you covered! 🙌 Our team has created a video summary of our article on evaluation of RAG Retrieval Chunking Methods. In just one minute, you'll discover insights and pro tips you need to boost your RAG retrieval performance. 📈 🔍 Key highlights: - The top-performing chunking method that outshined the competition by 10% - The surprising effectiveness of the TinyBERT reranker - Why sentences are the ideal units for chunking - Actionable strategies for maximizing your RAG retrieval pipeline Whether you're a seasoned professional or just starting out, this video is packed with valuable information to help you stay ahead of the curve. 🌟 📹 Watch the video: https://buff.ly/4b2CqtQ 📄 Read the full article: https://buff.ly/4aZt5TF #RAGRetrieval #DataScience #MachineLearning #Efficiency #Optimization #Superlinked

1

It's time for another part of our article series about #chunking methods 👉 https://lnkd.in/dPB2dAsF. Two weeks ago, our Data Science expert, Konrad Ruciński, explained what chunking methods are, navigated through their kinds (including the introduction of our own method), and why they are so important for #RAG optimization. If you missed it, catch up 👉 https://lnkd.in/dFawVgJv. This time, he evaluates the methods focusing on their #costs and impact on #quality to help you navigate the options and choose the right one for your RAG. So dive in, expand your knowledge and ensure the RAGs you will build are as cost-efficient and accurate as possible thanks to choosing the correct chunking method.   #LLM #datascience

15

In this Technical Deep Dive, we share our approaches in using Large Language Models (#LLMs) to make numerical-type data(sets) like DHS, JMP, HDX more accessible with the example of WASH data. In this 15 min. video Olivier Mills and team cover: 1️⃣ Basic scripting to extract core data from datasets from API endpoints 2️⃣ How to use LLMs to generate LLM-friendly schema definitions 3️⃣ How to structure a AI-friendly database store that can scale to thousands of datasets and millions of data indicators 4️⃣ How to use vector querying to retrieve relevant indicators from the store based off of a natural language (english) query 5️⃣ How to build a prompt that generates valid SQL query using LLMs with the above retrieved data schema. The various components form a single AI system combining data ingestion and data exploration using a software layer, a data layer and multiple LLM layers. ⏭️ Next steps will be to expand the datasets, automate other data source ingestion and further test the system with more complex plain english queries. We plan to incorporate this as a new feature for WASH AI after testing. Feel free to reach out if you would like a copy of the Notebooks, sample prompts and data schema definitions. We are #opensource The initial work was done in collaboration with Data for Good YYC (see https://lnkd.in/g3AijS93)

3

1 Kommentar

A Survey of RAG and RAU: Advancing Natural Language Processing with Retrieval-Augmented Language Models Quick read: https://lnkd.in/gnGqk44W Paper: https://lnkd.in/g8-VftbX #llms #largelanguagemodels #ai

20

RAG with Mistral AI? Here is a cookbook 😎 this is awesome, I really wan to see more work done to show how to use open source & open weight models. note that all of Mistral's models are open weight models. #generativeartificialintelligence #genai #rag

2

📣 New in MEAP! 📣 Knowledge Graph-Enhanced RAG, by Tomaz Bratanic and Oskar Hane. 📚 Have you ever considered incorporating structured data into your #RAG applications? While most people focus solely on unstructured data as the primary source for RAG, you can also utilize #structureddata. Our latest MEAP release, "Knowledge Graph-Enhanced RAG," provides practical techniques to help you build RAG using both unstructured and structured data. This book is a practical guide to empowering #LLMs with RAG. Throughout the book, you will learn how and why storing data as a graph can enhance RAG accuracy, performance, and traceability. You will also learn to build a vector similarity search retrieval tool, create an Agentic RAG application, extract information from text to form a #KnowledgeGraph, evaluate performance and accuracy, and much more. 🚀 Take action now! Save 50% with code libratanic350. 📖 Get into the book: https://mng.bz/4JNB 📹 Check out this video to find out: https://mng.bz/v87x #GraphRAG #ManningBooks #LearnwithManning

105

1 Kommentar

As for many people coming from academia, hyper-fixating on details and deep diving in a subject are second nature to me. But when I started to develop a product that needed to tackle real problems of clients, I had to learn a different approach: I started to think in term of use cases. The point is not to develop a perfect algorithm in absolute sense, but an algorithm that is good for the intended use. “Good enough” is better than “perfect”, when “good enough” can be turn into reality while “perfect” can live only as an ideal. This does not mean that the attention to details is useless: having an overview of what has been covered and what can be still improved is crucial, having an idea of what perfect would look like is greatly beneficial, but one has to learn to prioritise. Optimising your code for gaining one millisecond when the variability of the execution of your algorithm is of the order of seconds is not priority, but optimising your code to remove a bottle neck that will be reached imminently it is! This is a way of approaching problems that I keep learning every day. Here are few things, in random order, that help me in my personal development journey: - Run prioritisation sessions. Include people with different expertise in the session, the priority has to be set by inspecting different aspects, from backend to marketing. Surround yourself of people that challenge you to reconsider the priority you set and challenge others to defend their priorities. - Make sure that your code is easy to maintain and improve. What is good enough today, will not be good enough forever, you will need to iterate on your algorithm over and over.  - Take time to learn what others in your company are doing. Sometimes take on tasks outside your comfort zone/expertise: they help you seeing “the big picture”. - Keep track of the development already done and a backlog of the improvements that can be done. All the details, that we, academics, love so much and are so good to spot, should be documented to be picked up when appropriate. - Try to connect at least one use case to each improvement, this will help with the prioritisation. - Talk to clients, what are their needs? Always challenge yourself to explain your algorithms to non experts. The final user of the algorithm will most probably not be an expert and, if you want a useful feedback from them, they need to be able to understand the value of your work. Bonus wisdom: - Don’t lose your sleep if all the corner cases are not immediately covered. Also your product is on a journey to improve, investigate the limitations of your algorithm to be ready to improve it when needed. #DataScienceLife #AcademiaInTheWild

11

📢Remote Sensed and/or Global Datasets for Distributed #HydrologicalModelling: A Review by Muhammad Haris Ali, Ioana Popescu, Andreja Jonoski and Dimitri P. Solomatine https://lnkd.in/e9FEM4a3 #hydrology

32

2 Kommentare

Just wrapped up guest lecturing in STATS 352 at Stanford University on R & Python WebAssembly (WASM) backends for reproducible, interactive data science documents in the browser with Quarto! 🎉 Slides 📊: https://lnkd.in/eQ6ezcYE Repo 🔗: https://lnkd.in/emdj2JBH #quarto #webr #rstats #pyodide #python #datascience

73

1 Kommentar

"In this article, we discover the strikingly simple concept behind iTransformer and explore its architecture. Then, we apply the model in a small experiment and compare its performance to TSMixer, N-HiTS and PatchTST." iTransformer: The Latest Breakthrough in Time Series Forecasting by Marco Peixeiro

54

1 Kommentar

GraphRAG: New tool for complex data discovery now on GitHub GraphRAG, a graph-based approach to retrieval-augmented generation (RAG) enhancing question-answering over private or unseen datasets, is now available on GitHub. This tool utilizes a large language model (LLM) to construct a rich knowledge graph from text documents, identifying semantic structures before any queries. By detecting and summarizing densely connected "communities" of data, GraphRAG provides comprehensive overviews and enables accurate answers to global questions that naive RAG approaches cannot. Evaluation shows GraphRAG outperforms naive RAG in comprehensiveness, diversity, and empowerment of responses while being more token-efficient than traditional summarization methods. This release includes a solution accelerator on Azure for easy, code-free deployment. Read more here (Microsoft Microsoft Research): https://lnkd.in/eFwymjkV #LLM #GraphRAG #MachineLearning #KnowledgeGraphs #GitHub #Azure #RAG #Microsoft #DataScience #AI

1

"I will show you how we can easily implement a N-BEATS model in Python and also tune its hyperparameters." Dr.-Ing. Jonte Dancker shares an easy-to-understand deep dive into how N-BEATS works and how you can use it:

191

I'm excited to share the preprint "Pathological Regularization Regimes in Classification Tasks." In it, Max Wiesmann and I show how naturally occurring datasets can cause regularized machine learning algorithms to predict exactly the wrong trend. https://lnkd.in/eqneNMfu Our running example is credit scoring, where the synthetic data we use features the trend that female applicants for a loan are less likely to default in each occupation group compared to their male counterparts. For ridge regression, we give a precise characterization for which regularization parameters lead to the fitted model predicting the opposite trend, that female loan applicants are more likely to default than their male counterparts. For logistic regression, we give numerical evidence that all datasets satisfying Simpson's paradox exhibit the same trend reversal pathology. We also show that this trend-reversal pathology occurs in datasets already in the literature, including datasets from clinical trials, university admissions and death-penalty judgements along the race of the defendent and victim. Each of these cases, as well as our synthetic data about loan applications, falls under the "High Risk AI" category in the EU's AI Act regulation. #machinelearning #datajustice #AIACT #Regulation

21

Retrieval-Augmented Generation (RAG) enables answering functional, technical and domain-related questions about large codebases using LLMs, thus supporting developers in onboarding, feature development, or error analysis. In traditional RAG, code fragments relevant to the query are selected using similarity search. However, this method can miss critical code segments that are not directly related to the query but are crucial for correctly answering the question. As part of her master's thesis at itestra, Helena is researching an extension of RAG with enhanced code retrieval capabilities based on LLM agents. This aims to deliver more precise code snippets, significantly improving the quality of the answers. Helena uses Llama 3 70B and runs the entire setup on-premise on our own Nvidia RTX 6000 Ada, making it suitable for projects where the source code should not be processed in the cloud. #SoftwareEngineering #CodeAnalysis #ArtificialIntelligence #GenerativeAI #RetrievalAugmentedGeneration #LargeLanguageModel

27

4 Kommentare

I will be attending #ICML Monday-Thursday in Vienna, Austria. If you are around and want to chat about large-scale graph learning or sequence learning and how we use them in Meta ping me. #ICML #ICML2024 #Meta #GraphLearning #SequenceLearning

42