Tahrima Mustafa

When it comes to "Retrieval" phase of RAG, choosing the right strategy makes all the difference. Sparse retrieval (like BM25/TF-IDF) is great for finding exact keyword matches, while Dense retrieval (like BERT) understands the meaning behind words for more accurate results. Hybrid retrieval combines both, offering the best of both worlds. Each method has its strengths. Understanding when to use them is key to better search results. 

10

4 Comments

New Berkeley on understanding large language model: foundations and safety. Lectures on YouTube and covers below topics. https://lnkd.in/g5ukTUV2

9

📊 Optimizing RAG Applications with TruLens: A Quick Guide Recently explored TruLens for evaluating and fine-tuning Large Language Models (LLMs) in Retrieval Augmented Generation (RAG) applications. Here's what I learned: Key Evaluation Metrics: ✅ Groundedness: Ensures responses are factually supported by context ✅ Answer Relevance: Measures if responses directly address queries ✅ Context Relevance: Evaluates quality of retrieved information The Process: Set up automated evaluation pipelines using TruLens Track performance across different configurations Use metrics to optimize: Chunking strategies Retrieval parameters Context window sizes Embedding models Benefits: Data-driven optimization Automated quality assessment Systematic performance tracking Clear metrics for improvement Result: More accurate, relevant, and reliable RAG applications with measurable improvements in response quality. #AI #MachineLearning #LLM #RAG #DataScience #TruLens https://lnkd.in/gHQnjREi

3

Data aficionados, I've stumbled upon some incredible tools that are real game-changers in our industry. Here’s a quick rundown: 1. Vectorize is designed for Retrieval-Augmented Generation (RAG). It effortlessly converts unstructured data into high-performing vector search indexes. Imagine cutting through data chaos like never before. Do any of you folks already use this? 2. Spreadsheet is all you need—a mind-blowing nanoGPT pipeline neatly packed in a spreadsheet format. Perfect for anyone looking to get a handle on GPT's intricacies without diving into hardcore programming. 3. Replicate makes it a breeze to run and fine-tune open-sourced AI models using a straightforward API. Talk about simplifying complex operations! What AI models are you all experimenting with right now? 4. transformers.js brings the power of transformers right to your browser. Think about the possibilities here—AI execution without ever needing to leave your browser. Have you guys tried this yet? 5. Build your own X is a treasure trove of step-by-step guides for creating your own tech marvels from scratch. Whether it's AR, bots, or torrent tech, this resource has got you covered. Curious, which tech are you excited to recreate? 🚀 It's an exhilarating time to be in the data industry with tools like these at our fingertips. Let's elevate our game! #DataScience #AI #MachineLearning #VectorSearch #GPT #TechInnovation #DIYTech #Transformers Jump in and share your experiences or thoughts about these tools! 🌟

5

1 Comment

🧠 For my folks who are interested in learning more about programming in R: Harvard's CS50 is offering a free course on Introduction to Programming with R on edX! This is a great FREE opportunity to level up your data skills in a powerful language for statistics and data visualization. 📈 Whether you're new to coding or looking to add R to your toolkit, this self-paced course is a great place to start! #CS50 #RProgramming #DataScience #LifeLongLearner

2

Recently I have been using cmeans and topic modeling more and more. I am a big fan of non-exclusive clustering methods especially when using the results of those approaches as features for another model. I just find it must less restrictive than exclusive clustering methods like kmeans and often yield model results inprovement. https://lnkd.in/eVygG35z

25

2 Comments

Data Scientists, This is BIG! Snowflake's Arctic LLM is Here – and Open-Source! 🤯 Get your hands on it now: Interactive Demo: Explore Arctic's capabilities with this Streamlit app → https://lnkd.in/gJ7y38Ng Replicate: Ready to experiment? Find it on Replicate → https://lnkd.in/gyDXiMju Why This Matters: Arctic lets you quickly build demos of your models. This is a game-changer for showcasing your work and making your models more accessible! Huge thanks to Snowflake for making this happen! Are you excited to try Arctic? Let me know in the comments! 👇 #LLM #OpenSource #DataScience #Streamlit #Snowflake

8

We make decisions weighing out uncertainties and risks all the time… we need models to present their outputs with uncertainties to make fully informed decisions as we use them

7

Exploring non-linear relationships in regression? Transformations can lead to trillions of potential models, but is there a better way? Samuele Mazzanti's article discusses how gradient-boosting models can provide effective solutions with ease. #Regression #DataScience

8

1 Comment

Great keynote at the AirFlow Summit today by Oliver Dykstra about data engineering and analytics at the Texas Rangers ⚾️. For those who saw Moneyball, baseball is a data-driven sport, and it is amazing to see the level of analytics used to improve the team's performance 🚀. I am not a baseball fan, but after this talk, it definitely sounds much more interesting 😅! #airflow #data #dataengineering #datascience

50

3 Comments

A clean (devoid of unnecessary abstractions) experimental package released by OpenAI for multiagent systems. There is no official support as of now to the package, but few interesting examples. Also like their simple take on defining agents - An Agent simply encapsulates a set of instructions with a set of functions (plus some additional settings below), and has the capability to hand off execution to another Agent. MIT license. Good read for exploration. Link in comments. [openAI, agents, LLMs]

11

2 Comments

Check out this highly recommended course for those interested in optimizing their models: - Introduction to Quantization - Symmetric and Asymmetric Quantization Modes - Hands-on Implementation from Scratch using PyTorch - covering Per Tensor, Per Channel (Axis), and Per Group techniques - Weights Packing: Exploring 2 or 4 Bits Quantization Methods https://lnkd.in/gMAuWXHm

43

Pretty good short course, containing information about dense and sparse retrieval, that might be useful for understanding contextual retrieval approach, evaluation using Ragas, and the Graph RAG.

3

We had this coming: MoA LLM -> Mixture of Experts (MoE) Agents -> #Mixture of Agents (MoA) Let's understand them in simple terms. #MoE: Within one large model, sub-models are created on specific parts of data to develop an expert on that part of data. E.g. One sub-model specializes in predicting health conditions for Native Hawaiians while another sub-model is specialized for Asians. MoE reduces training time, computational costs while maintaining (or improving) accuracies of #LLMs. Twitter / X's #Grok-1, Mistral AI's #Mixtral 8x7B (8 experts of 7 billion parameters each), Databricks's #DBRX are based on this architecture. Ensemble models (EM): Why am I suddenly jumping to EM? Because, in your discussions, this may come up as a concept similar to MoE at #LLM-level, just like EM is at traditional #ML-level (e.g. Gradient Boosting Random Forest). But there is one distinction - In EM, each model is trained on all the data while in MoE, each sub-model (expert) is trained on a subset of data. #MoA: While MoE combines multiple neutral networks (#experts), MoA combines outputs from multiple Agents (potentially accessing multiple LLMs internally). Advantage: More accurate response with improved quality. Disadvantage: Longer inference time; to the time order of 5x. Solution: Groq. It reduces inference time to as low as 1x! #ArtificialIntelligence #AI

20

Mining Online Data Across Social Networks & Internet Data Capturing Data, Modeling Patterns, Predicting Behavior. Capturing Data, Modeling Patterns, Predicting Behaviour... Capturing Data, Modeling Patterns, Predicting Behavior. Capturing Data, Modeling Patterns, Predicting Behavior - Based on collecting more than 20 million blog posts and news media articles per day, Professor Jure Leskovec discusses how to mine such data to capture and model temporal patterns in the news over a daily time-scale. In particular, the succession of story lines that evolve and compete for attention. He discusses models to quantify the influence of individual media sites on the popularity of news stories and algorithms for inferring hidden networks of information flow. Topics : 0:00​ Introduction 0:08​ Meet The Speaker 4:01​ Many data is a Network! 5:27​ Networks: Rich Social Data 6:27​ Networks: Size *matters 7:38​ Networks: Structure & Process 9:02​ Why study Web and networks? 10:28​ Projects: Link Prediction 17:17​ Projects: Friends vs. Foes 18:13​ Projects: Predicting Enemies 20:26​ Theory of Structural Balance 22:25​ Theory of Status 24:18​ Friends vs. Foes, Trust vs. Distrust 27:55​ Online (social) media 29:50​ Tracking Information on the Web 32:29​ How is news being made? 34:35​ Modeling influence 36:41​ Influence curves of media types 38:35​ Inferring the Diffusion Network 39:26​ Inferring networks 42:21​ Maximizing the Influence 43:35​ Influential blogs & Information outbreaks 44:29​ Reflections 46:03​ Directions 47:23​ References 47:34​ Autumn Quarter 2011-12 49:09​ Graduate Portfolio 50:04​ Course Enrollment 51:06​ Please indicate your level of interest in the Mining Massive Datasets graduate certificate upon registration. Register for this online video webinar session to know more. https://lnkd.in/gKnizreE

1

There is often an unspoken assumption by ML researchers that all we need to make progress and model every aspect of our world is data. We do not ascribe to this in my lab. In fact, we believe that data will never be enough. Our experiences with the nuances and complexities of scientific problems have informed us to the insufficiency of data to capture continuous physical processes, which afterall govern our biological and physical world. An example of this is this series of two papers led by my wonderful PhD student, Anowarul Kabir and advanced by a precious multi-year collaboration of my lab with Los Alamos National Lab: Anowarul Kabir, Manish Bhattarai, Kim Rasmussen, Amarda Shehu, Anny Usheva, Alan R Bishop, and Boian S Alexandrov. Examining DNA Breathing with pyDNA-EPBD. Bioinformatics 39(11):btad699, 2023. https://lnkd.in/gZaHE4hS Anowarul Kabir, Manish Bhattarai, Selma Peterson, Yonatan Najman-Licht, Kim Ø Rasmussen, Amarda Shehu, Alan R Bishop, Boian Alexandrov, Anny Usheva. DNA Breathing Integration with Deep learning Foundational Model Advances Genome-wide Binding Prediction of Human Transcription Factors. Nucleic Acids Research: gkae783, 2024. https://lnkd.in/gfTTqcPy Our goal: advance an exceptionally challenging problem in molecular biology, prediction of transcription factor binding sites. Our first step: capture the underlying physics that is missing in the data. Our second step: integrate that now with the data we have in a foundation model for predicting transcription factor binding sites. Performance improves. Most importantly, when we look at the sequence motifs that constitute "signatures" of what makes a transcription factor binding site, we obtain answers. All in the open, nothing opaque.

33

1 Comment

In the tenth part of this article series, we discuss the role of time horizons and discounting in Markov Decision Process. #data #analytics #ai #ml #datascience #designedanalytics #markovprocess #markovdecisionprocess #reinforcementlearning #deeplearning #deepreinforcementlearning https://lnkd.in/gXbvnBHT

1

1 Comment