Aishwarya Naresh Reganti的动态

Extracting insights and topics from text data can be challenging. I describe a method using LLMs to streamline dataset curation for supervised learning models. This approach speeds up the labeling process, ensuring that the generated topics are relevant and meaningful. Check out the post here: https://lnkd.in/gvAsi83Y. Dive into the repository for more details: https://lnkd.in/gWpfcNix.

What are #LLM trained on? On big corpuses of text and data. For example, on the Common Crawl dataset. I found this article pretty interesting to better understand how such datasets are prepared, filtered, and cleaned, before really been in a good shape for ingestion during the model's training and learning process. https://lnkd.in/efUDmZG7

January Is for Challenging Yourself to Learn New Skills: You can start your data science journey at any time; expanding your skill set should be an ongoing, yearlong process. Still, even those of us who are skeptical of new year’s resolutions can’t deny the sense of excitement and opportunity that comes with a whole, blank-slate year on the horizon. What better time to take the plunge and explore new?topics? To give you a helpful nudge in that direction, we’ve put together a lineup of fantastic articles from recent weeks that focus on accessible, practical approaches to machine learning and data workflows. Many of these are beginner-friendly, but as we often remind ourselves: you’re always a beginner when you decide to learn something new. We hope you enjoy our selection this week, and that it inspires you to take on new challenges throughout the year. Let’s dive?in. * Courage to Learn ML: A Detailed Exploration of Gradient Descent and Popular Optimizers In a new installement of her series of helpful machine learning explainers, Amy Ma offers a thorough and accessible guide to gradient descent and other optimizers, and focuses on choosing the right one depending on the task you’re aiming to complete. * From Adaline to Multilayer Neural Networks If you feel like you’re not entirely on firm footing when it comes to all those complicated mathematical notations in machine learning papers, Pan Cretan’s latest deep dive is an excellent resource. It goes back to the early days of multilayer neural networks, builds one from scratch, and unpacks these networks’ mathematical descriptions. * A Comprehensive Overview of Gaussian Splatting If you’re a more advanced practitioner who likes staying up-to-date with recent research, Kate Yurkova’s primer on Gaussian splatting is a must-read. It’s an ideal starting point for exploring this emerging approach for 3D representation and its various real-world use?cases. Photo by Jeff Sheldon on?Unsplash * LLM Agents?—?Intuitively and Exhaustively Explained Ready to roll up your sleeves and start tinkering with large language models? Daniel Warfield recently shared an exhaustive, one-stop resource on LLM agents, how they work, and how they can help you streamline your interactions with AI?tools. * A Bird’s Eye View of Linear Algebra: Systems of Equations, Linear Regression and Neural Networks It’s never too late to bolster your math foundations or refresh your knowledge of concepts you only learned about in passing or a long time ago. The latest post in Rohit Pandey’s ongoing series on linear algebra takes a close look at solving systems of linear equations. Still feeling energized? We hope so?—?here are a few other notable contributions we wanted to share with you this?week: * Diving deep into the problem of parameter efficiency, Mariano Kamp offers a detailed analysis of the design decisions that… #MachineLearning #ArtificialIntelligence #DataScience

Combining LLM-generated knowledge graphs and graph machine learning, GraphRAG will be more poplular in the near future. Graph Vector embeddings enable the information in the form of What, When, Who, and Where. Knowledge graphs on the other hand represent the reason and deep interlinking of various information.?

"Exploring the world of data processing tools! ?? This captivating image showcases operations in Pandas, Polars, SQL, and PySpark, offering a glimpse into the diverse landscape of data manipulation. From the intuitive interface of Pandas to the blazing fast performance of Polars, and the structured querying power of SQL to the distributed computing prowess of PySpark, each tool brings its own strengths to the table. As data enthusiasts, it's essential to navigate these options wisely to unlock insights efficiently and effectively. Which tool do you find most compelling for your data projects? Let's spark a conversation! #AI #Innovation #MachineLearning #DataAnalytic

pub.towardsai.net: The article discusses the importance of Pandas for data scientists and encourages readers to refer to a previous post for more information. It invites readers to continue reading on Towards AI for further insights.

Did you know you could speed up fine-tuning with dataset thinning? As you know, fine-tuning large language models on massive datasets is time-intensive & expensive. Which in turn makes quick iteration and model refinement a challenge as every training cycle takes up a lot of time. To handle this challenge, we came up with a way you can thin your dataset while maintaining the quality. How? Let's learn Step-by-step breakdown of Clustering a Dataset & Thinning Redundancies. Step 1: Install Required Packages & Load the Dataset Step 2: Choose an Embedding Model Step 3: Embed the Dataset & Apply DBSCAN for Clustering Step 4: Retrieve Cluster Assignments & Visualize Cluster Distribution Step 5: Thin Redundant Clusters & Fine-tune your model Read in detail how it works in this latest guide - https://lnkd.in/gJKYaB6j

In this insightful blog post, we delve into the fascinating world of decision trees in machine learning. From predicting stock prices to diagnosing diseases, decision trees are an essential tool in data science. Whether you're a seasoned data scientist or a curious beginner, this post offers valuable insights into this powerful machine learning technique.

At :probabl., ?? Vincent Warmerdam has been creating a bunch of really good content related to ML (non-hype-y, calm!, and useful ones ?? ). I've already encountered a bunch of cases where people talk about their struggles and issues at work dealing with data, and some of these videos very nicely tackle those issues. I particularly love the "Whiteboard Series" where he goes into a 10-12min dive of a certain topic. For all those videos, we have a discussion internally before publishing them; so you know a bunch of scikit-learn maintainers, the likes of Olivier Grisel and Guillaume Lemaitre have already vetted them. I very much recommend having a look, if you use statistical models in your daily work. ?? https://lnkd.in/e2baNhRr ??

Just completed DevCoach 155: Data Science | Understanding Machine Learning Algorithms for Data Science from Dicoding Event on Friday, June 14, 2024. In this discussion, I learned various things such as the definition of machine learning, the relationship between machine learning and data science, the introduction of deep learning algorithms, and more. I am very excited apply this knowledge into my skillset.