Segun Umoru

Evaluating the Performance of RAG Applications 🚀: I've noticed a trend recently around the challenges of effectively evaluating RAG performance. Just today I’ve come across three separate posts on this topic. There are numerous lenses to consider, should we focus on the retrieval process, the generated answer, or a combination of both? What metrics should we use, and are there any established frameworks to help us? Through my own research and exploration, I've identified a few specific challenges. Firstly, the existing frameworks begin to break down if you plan to use an LLM as a judge and don't want to use an OpenAI-flavored model. Secondly, there seems to be a lack of modularity in regard to evaluating my specific data with the metrics I care about. After experimenting with several libraries, I've narrowed it down to two primary options: Arize AI and ZenML. If built-in evaluations are your preference, Arize AI offers curated datasets, and a wide range of telemetry to support your orchestrator. On the other hand, if modularity is your priority, ZenML provides you with the observability you need without the abstraction, allowing you to evaluate your application in the way that best suits your requirements. Both supply a visually appealing UI. As always, I'm interested to hear how your thoughts / how you've solved this yourself in the comments! Hamza Tahir Adam Probst ZenML Notebook -> https://lnkd.in/eM4FNKbe #openai #llms #datascience #rag #machinelearning #whatinspiresme #langchain #genai #artificialintelligence #evaluation

17

Creating Data Frames: Create a Data Frame from a NumPy array or a dictionary. Setting Index: Set the index of a Data Frame to a list of labels. Getting Values: Get the values in a Data Frame as a NumPy array using the .values attribute. Extracting Columns: Extract one or more columns by their position using .iloc or by their header label using square bracket notation. Getting Rows: Retrieve the first 5 rows using .head() and the last 5 rows using .tail(). Sampling: Sample a random row using the .sample() method. Column Operations: Set column names using a list of labels and check the data types of columns using the .dtypes attribute. Generating Statistics: Generate summary statistics using methods like .describe(), .mean(), and .count(). Extracting Rows: Extract rows by their position using .iloc or by their index label using .loc. Modifying Data: Add new columns using assignment or .insert() and add new rows using .append(). #Pandas #Datascience

41

Very interesting module on Data Modelling- especially how to use PBI to execute dimensional modelling and making Star and Snowflake schemas. Performance comparison between the types of schemas was a real eye opener.

1

https://lnkd.in/gAK3n_bD 🚀 Exciting News: Free Access to Euron's Courses! 🚀 Hi everyone! Euron, a new learning platform, is offering free access to courses across a variety of tech domains for a limited time! Now’s your chance to upskill, learn something new, and explore tech topics at no cost. 🎓 How to Get Access: Sign-Up Link: https://lnkd.in/gJehgymH Google Form: https://lnkd.in/gpMvgsGK For Queries: Join group: https://lnkd.in/g8NDg8ka Check Out Learning Logic YouTube Channel: https://lnkd.in/gtrXrReE Feel free to share these links with anyone who might be interested. Don't miss out on this opportunity to get ahead in tech for free! #AIDangers, #AIThreats, #DeepfakeAwareness, #AIandPrivacy, #AIEthics, #JobDisplacement, #AIWeaponization, #AIBias, #AIHealthcare, #ArtificialIntelligence, #ResponsibleAI, #AI2025, #Deepfakes, #SurveillanceConcerns, #EnvironmentAndAI, #AIInCreativeFields, #EthicalAI, #FutureOfAI, #AIRegulation, #TechRisks

14

1 Comment

Data visualization is very important part in of Data Analysis. The benefits of data visualisation are 1. Accessibility: since Information is easy-to-understand for a variety of audiences. 2. Visualize relationships and patters between data when the information is presented in a graph or chart. 3. Explore the the of collaboration between data so as to make decisions.

1

Simplify the creation of publication-ready plots! The ggpubr package builds on ggplot2, providing easy-to-use functions for creating clean and stylish visualizations, perfect for reports, articles, and presentations. ✔️ Publication-Ready Graphics: Designed to help you generate high-quality plots with minimal effort, ensuring your visuals are polished and professional. ✔️ Easy Statistical Annotations: Easily add statistical comparisons, p-values, and significance markers to your plots, making it simple to highlight key differences. ✔️ Streamlined Plot Customization: Includes intuitive functions for adding titles, labels, themes, and more, so you can quickly fine-tune your plots to match your style. ✔️ Seamless Integration with ggplot2: Works directly with ggplot2, allowing you to enhance and customize your plots without learning a new syntax. The visualizations shown here are from the package website, demonstrating how ggpubr can simplify the creation of beautiful, informative plots: https://lnkd.in/e9ZbAyt2 Want to master ggplot2 and its powerful extensions to create stunning visualizations? Enroll in my online course, “Data Visualization in R Using ggplot2 & Friends,” starting on November 25, 2024! Take a look here for more details: https://lnkd.in/ehRrzdJ7 #dataanalytic #datavisualization #rprogramming #coding #tidyverse #visualanalytics

56

📢 Reminder: Final Python Basic Coding Session Tomorrow!  Hello, everyone! 👋  Our free 3-week Python Basic Coding Sessions led by our amazing tech tutor, Patric Olumba, are coming to an end tomorrow, Saturday, 7th December, at 12 noon. Let’s make it count!   Tech group: https://lnkd.in/dsxz6H5g The live class will be on Google Meet. Here's a recap:   ✅ Week 1: Introduction, installation, and basic syntax.   ✅ Week 2: Syntax, indentation, data types, and operators.   ⏳ Week 3 (Tomorrow): Variables, data types, and operations.  Don't miss this final chance to level up!  This is your last opportunity this year to gain these foundational coding skills. Let’s show up on time and not keep Patric waiting.  See you there! Let’s finish strong! 💻🔥 #python #coding #education #WorldComIntellectuals

2

Brushing up on some python skills to develop a skills checker. ✅ Have a look at the code here https://lnkd.in/ercsyb3g I want to add this to an interactive website. Does anyone have any experience of Flask and can recommend some helpful resources? 🤔 #ContinuousLearning #Python #CodeFirstGirls #Flask

3

I built a text classification model using an LSTM,GRU , Bidirectional RNN/GRU RNN network with TensorFlow. I worked with the IMDb dataset to train the model on analyzing movie reviews. The process involved preparing the text data, converting words to numerical values, and training the model to predict sentiment classifications. This project was a great opportunity to apply deep learning techniques in Natural Language Processing. Topics: #Text_Preprocessing #Data_Handling #Feature_Engineering: Models: GRU (Gated Recurrent Unit): Bidirectional RNN/GRU RNN DataCamp DataAnalyst.com #NLP https://lnkd.in/d2DVpNcq

52

7 Comments

I've recently published an article titled "Introduction to the R Programming Language – Basic Concepts for Beginners." In this article, I cover fundamental features of R and explain how to use it to manipulate, analyze, and visualize data. The article was created in collaboration with Tomi Mester, and you can find it on the Data36 website: https://lnkd.in/ejZGsVUn Here’s a glimpse of what you’ll learn: 1️⃣ Background information about the R programming language. 2️⃣ Data manipulation techniques in R. 3️⃣ Calculating descriptive statistics for your data sets. 4️⃣ Creating a correlation matrix to understand variable relationships. 5️⃣ Estimating a linear regression model. 6️⃣ Visualizing data using the ggplot2 package. You can easily follow along by copying and pasting the R code snippets provided into R or RStudio. Every code example is fully reproducible. If you're eager to dive deeper into R, check out my course, "Introduction to R Programming for Absolute Beginners." This course is perfect whether you're starting from scratch or looking to refresh your skills. Check out this link for more details: https://lnkd.in/ewKXsXwJ #rprogramming #ggplot2 #database #package #dataviz #research

48

Day 27: Transfer Learning Transfer learning leverages pre-trained models to solve new, related problems, allowing you to use existing knowledge for more efficient training on new tasks. It’s widely used in image and language tasks to speed up training and improve performance, especially when data is limited. Key Concepts: 1. Pre-trained Models: Models like VGG, ResNet, and BERT trained on large datasets. 2. Fine-tuning: Adapting the model’s learned features to a new task. 3. Feature Extraction: Using existing model layers to process new data. Why Use Transfer Learning? Reduces training time and data requirements. Improves accuracy, especially in complex problems. Enables tackling tasks with limited data. #MachineLearning #TransferLearning #DeepLearning #AI #DataScience #PretrainedModels

2

K-Nearest Neighbours using classical Iris Flower Dataset Today, machine learning is one of the most buzzed-about and in-demand fields. Before diving into complex algorithms, it's crucial to conceptually understand the basics, say, simple classifiers like K-Nearest Neighbors (KNN). Understanding the concepts behind algorithms like KNN is crucial. It enables us to make informed decisions about data preprocessing, parameter tuning, and validation. Coding, on the other hand, brings these concepts to life. It allows us to apply theoretical knowledge to real-world datasets, uncover patterns, and draw meaningful insights. Together, conceptual understanding and coding proficiency form a powerful toolkit for any analyst or data scientist. Excited to share our latest project on K-Nearest Neighbors (KNN) using the Famous Iris flower data set! In this comprehensive analysis, we've covered everything from data cleaning and exploratory data analysis (EDA) to a detailed description of the dataset and an in-depth explanation of the KNN methodology and validation process. All tasks were executed in Python to ensure precision and reproducibility. Check out the full repository on GitHub to explore the code and insights: https://lnkd.in/gy8vdzwD #DataScience #MachineLearning #Python #KNN #DataCleaning #EDA #GitHub #AI #Tech

15

1 Comment

Dynamic Conditional Formatting with Radio Buttons: Part 2 (Full Tutorial) In this second part, I complete the full tutorial on creating dynamic data views with radio buttons. Now you can fully implement this interactive conditional formatting technique to make your data analysis smoother and more engaging. If anything is unclear, feel free to ask in the comments. 👉 Follow me, Shristi Kumar for more Excel tutorials and tips! #alignexcel #excelproductivity #excelvisualization

11

🚀 I've just published a new article on "Structured Data Generation"! 🌟 If you're interested in the best approaches for generating structured output, don't miss my detailed comparison of methodologies.💡 Your feedback and thoughts are highly appreciated! Let's discuss and learn together! 📚💬 Thanks Vaibhav for Proofreading 🤝 #LargeLanguageModels #BAML #StructuredData #DataScience #AI #MachineLearning #OpenAI #JSONParsing #TechInsights #LangChain #Kor #Instructor #Mistral #PromptEngineering #JSONMode #FunctionCalling #JSONFixing #Outlines #GuardrailsAI #Guidance

17

2 Comments

Final reminder: the Statistics Globe online course, "Data Visualization in R Using ggplot2 & Friends," starts today! It would be great if you also took part in the course. So if you are interested, please register now: https://lnkd.in/ehRrzdJ7 #rstats #ggplot2 #statistics #datascience #dataviz

44

4 Comments

What is RAG? RAG stands for Retrieval-Augmented Generation. It is a technique that enhances the capabilities of language models by combining two components: Retrieval: A search or retrieval component that fetches relevant documents or data from external sources (e.g., databases/vector databases, knowledge bases, or the web) in response to a query. Generation: A generative model (like GPT) that processes the retrieved information and generates a response. By retrieving relevant information and then using a generative model to synthesize a coherent and contextually appropriate answer, RAG helps in improving the quality and accuracy of responses. Example: Chatbot augmented on organisation data #generativeai #rag #chatgpt #gpt #ai #llm

12

certificates can Help You Get Hired??? #datascience #dataanalytics #businessanalytics #datavisualization #data #powerbi #tableau #statistics #artificialintelligence #machinelearning #deeplearning #jobs #career #jobsearch #reelsinstagram #uk #usa

13