Afiba Annor

Discover how eBay created their own language model using three billion item titles in this exciting video. We'll explore the innovative techniques eBay used, including training their BERT model from scratch on eBay's massive dataset, knowledge distillation to compress the model, and fine-tuning the compressed model to learn similarity better. With a 3.5% increased purchase order rate and enhanced customer engagement, this is a must-watch for anyone interested in the power of language models. Like and subscribe for more such interesting concepts. Also, like and share over here for maximum reach. : ) Video Link: youtu.be/h51nbWr7feo YT channel Link:  youtube.com/@datatrek #datatrek #datascience #machinelearning #statistics #deeplearning #ai

9

Had a great time at #ODSCEurope2024 this week, swapping stories with so many other #data, trad #machinelearning, and #generativeAI nerds like me. I want to elevate some brilliant women of data science who were there, too! Camilla Bressan, PhD, Hannah Ogden, and Regeane Bagonyi. These incredibly accomplished humans make me proud to be a part of the #S2DS network. #DiversityMatters #DataScience #AI https://lnkd.in/eKXPXqPf

17

1 Comment

LLMs are getting the bulk of LinkedIn discourse targeted towards senior and junior data scientists alike. This gives the impression that being skilled in language modeling is of critical importance to data science practitioners. However, if you work in a large enterprise or for a consultancy, you will inevitably have to spend your time developing other skills not related to language modeling. I think it is particularly salient to remember that tabular data still underpins the vast majority of operational, transactional, and customer data across many domains. Spend more of your time on LinkedIn following people who discuss statistics and the practical considerations of building machine learning models.

7

1 Comment

Following on from last week's article, this week I cover Activation Functions

1

I recently posted a few posts about Rust 🦀 and my intention to leverage it for data science applications. Multiple people asked if Rust is a substitute for R or Python, and the short answer (in my opinion) is no. I see Rust as a complementary or supporting language that could make languages like R and Python faster. Polaris 🐻‍❄️ is one example of a Python 🐍 application that uses Rust on the backend. I am mainly interested in the interaction between Rust and WebAssembly, and MLOps applications. Here is a great lecture by Lincoln Colling, PhD from the University of Sussex, about the data science applications of Rust 👇🏼 📽️ https://lnkd.in/dhGH_UBA #rust #datascience #python #rstats #mlops

64

5 Comments

🌟 Exciting blog post from NHS England on Reproducible Analytical Pipelines  🥁 Keep banging the drum Sam Hollings 📈 NHS England are sharing their journey on Reproducible Analytical Pipelines (RAP) and how it's enhancing the quality of their data processes. 🔍 What is RAP? RAP aims to automate statistical and analytical processes, making data analysis more efficient, transparent, and robust. This means less manual work and more time for data analysts to focus on what truly matters - delivering insights that improve healthcare services. 🌐 Why It Matters for the Culture of Data Services? Adopting RAP aligns the commitment to enhancing the quality and reliability of data services. It's not just about technology; it's about fostering a culture where continuous improvement and collaboration are at the forefront. 🛠️ Progress NHS England Data teams are integrating RAP across the organisation and It's still early days, but the impact is tangible and transferable. 📈 Looking Ahead They are committed to expanding the use of RAP, enhancing data capabilities, and continuing to share their progress. Hoping to hear more of this type of work. #NHS #DataScience #Innovation #RAP #HealthcareTechnology #Teamwork #Data #Health #Collaboration Link to the source in the comments. Image created by DALLE-3

17

1 Comment

Thanks to David Hoyle at dunnhumby and the Data Science Festival for a great session on demand modelling in the grocery retail industry! 🛒📊 🔍 Key Topics Covered: 1. Direct and Cross Price Elasticity: Understanding how changes in prices impact demand and how related products affect each other’s sales. 2. Practical Applications of Demand Models: Real-world scenarios where demand modelling plays a crucial role in decision-making. 3. Behind the Scenes: An introduction into the datasets, features, and models used to create accurate and insightful demand models.

20

1 Comment

Wanted to give a shout-out to Professor Kerby Shedden from the University of Michigan, who has written a fantastic 'Introduction to Data Science' ebook. I stumbled across this when searching for a good explanation of stratification to share with colleagues. It was extremely refreshing to find a book that begins by speaking to the foundations on which the field is built, rather than specific languages or tools. Don't get me wrong - I like Python and program in it regularly - but I loved that Python is mentioned only three times throughout the book, as one of many tools available. Too many intro's to DS start by launching into a language, and the worst of them closely follow that with specific sophisticated ML algorithms. These things belong later in learning about Data Science. Shedden doesn't cover the entire field and isn't trying to. His introduction is concise and well-written. I'd highly recommend this to anyone seeking a great summary of the foundations. #datascience

22

This week I learnt many interesting concepts about data predictions. The most interesting one gotta be Clustering and missing data. StatQuest explained well about random tree and random forests, where trees are used to build forest. Decision Trees: A Decision Tree splits data into branches based on leaves values to make predictions. Random Forest: A Random Forest combines multiple Trees to improve prediction accuracy. It uses random subsets of data to build each tree. In that video it explains the randomness of choosing those data testing sets. Bootstrapping: Bootstrapping involves randomly selecting subsets of the data to create those trees for the forest. Bagging: Bagging (Bootstrap Aggregating) combines the predictions from multiple trees to get a final result, making the model more accurate. So ... Decision Trees make predictions by looking at the leaves. Random Forests use many trees to increase high accuracy. Bootstrapping helps create diverse trees leaves samples. Bagging combines predictions from those trees for improved accuracy. 🙂 https://lnkd.in/eh56NfBZ

1

I’ve just received my copy of "Implementing Statistics with Python", a book I had the pleasure of reviewing. It’s an excellent resource for those looking to apply statistical methods using Python. Accessible and engaging, this introductory book is a pleasure to read https://lnkd.in/eh2knrSj. Many thanks to BPB Publications for the copy. #python #statistics #ai #machineLearning

44

11 Comments

LlamaIndex + MLflow RAG contains a lot of parameters to tune, from chunking to indexing - everything affects downstream answer accuracy. It’s important to have a systematic approach to track and tune these parameters, and define the right eval metrics/dataset to evaluate changes. Check out Jino Rohit's video on experimenting with a LlamaIndex pipeline with MLflow for experiment tracking. https://lnkd.in/gyYXfyyb

554

14 Comments

🚩 Did I Cram Python as a Data Scientist? Here’s the Honest Truth… You: “Favour, did you cram python”? Me: NO, because cramming doesn’t work when it comes to coding. 📍 If you’re aiming to become a data scientist, machine learning engineer, or software developer, one thing is for sure, you’ll be coding daily. Which means you can’t cram coding. 📍 Why Cramming Doesn’t Work How many methods, classes, or syntax rules can you memorize? Coding is not about memorization, it’s understanding why you’re using specific syntax or methods to solve problems. 📍 The ChatGPT Trap You cannot rely on ChatGPT to do your coding for you. I’ve experimented with it myself, and while it’s useful for quick snippets, about 85% of the time it gives errors. If you don’t know how to code, you’ll end up stuck in a cycle of copying, pasting, and fixing errors without even knowing what went wrong. 📍 How to Learn Coding Without Cramming Here’s what has worked for me, and it can work for you too: 1. Pick a Language: – Start with one programming language. Don’t try to learn multiple languages at once. 2. Start with the Basics: – Learn data types, variables, control structures, and functions. Don’t skip the foundations. 3. Break Problems Down: – When faced with a big coding challenge, break it into smaller tasks. Solve one piece at a time. 4. Understand the Why: – Don’t just memorize syntax; understand why you’re using certain methods or functions for specific problems. Once you grasp the “why,” it becomes easier to apply those concepts elsewhere. 5. Practice, Practice, Practice: – Build projects. Code regularly. The more you practice, the more you’ll improve. Each time you code, you’ll encounter new challenges, and that’s how you grow. 6. Use Resources, Don’t Rely on Them: – Tools like ChatGPT and Stack Overflow are great resources, but you need to understand what you’re copying and pasting. Otherwise, you’ll never truly learn how to code. 7. Pair Programming: – Pair programming is a fantastic way to learn. Working side by side with another developer allows you to see how they approach problems, and you’ll learn new techniques. If you’ve crammed, it’ll show. But if you understand the concepts, you’ll be able to keep up and contribute. 📌Note: AI should be a tool that supports you, not your main coding resource. The only way it can support you is if you already know what you’re doing. ♻️Repost so others can learn #favouribude #dataliving

770

146 Comments

Struggling to get your ML models from development to production? Feeling the pain of siloed data, inefficient deployments, and unclear model interpretability? You're not alone. Many data scientists face challenges in the critical MLOps stage. Abundent Academy's 'Mastering MLOps: From Fundamentals to Production' course can change that. This comprehensive program equips you with the essential knowledge and hands-on skills to: ✅ Build and manage efficient ML pipelines ️ ✅ Deploy models seamlessly to production on cloud platforms (think Azure!) ✅ Master CI/CD practices and leverage Kubeflow Pipelines ✅ Ensure model quality through performance auditing and data augmentation No prior MLOps experience? No problem! ‍ 👍 This course starts with the basics, taking you on a journey through the ML lifecycle and setting the stage for success. Ready to unlock the full potential of your ML models? Enroll today and become an MLOps pro! Spots are limited, so don't wait! **This course is HRD Corp claimable** Enroll here 👉 https://lnkd.in/gi525Fr8 🔵 Abundent Academy - Malaysia's Leading Automation & AI Academy #MLOps #MachineLearning #DataScience #AbundentAcademy #AI #HRDCorp #MalaysiaUpskill #Upskilling #Reskilling #FutureOfWork

5

Tired of the endless cycle of software installations just to get started with football data analytics? There's a better way! In my latest Substack post, I've laid out a smoother path to empower your learning journey. Discover lightweight quick alternatives to traditional tools that let you focus on what truly matters—honing your skills and understanding the game on a deeper level. Say goodbye to the installation blues and hello to insightful analysis. https://lnkd.in/ecJxq4yV #DataScience #FootballAnalytics #Efficiency

22

2 Comments

Discover how eBay created their own language model using three billion item titles in this exciting video. We'll explore the innovative techniques eBay used, including training their BERT model from scratch on eBay's massive dataset, knowledge distillation to compress the model, and fine-tuning the compressed model to learn similarity better. With a 3.5% increased purchase order rate and enhanced customer engagement, this is a must-watch for anyone interested in the power of language models. Like and subscribe for more such interesting concepts. Also, like and share over here for maximum reach. : ) Video Link: youtu.be/h51nbWr7feo YT channel Link:  youtube.com/@datatrek #datatrek #datascience #machinelearning #statistics #deeplearning #ai

10

Data Science Bootcamps are bad. Here's why they get things ass backwards: They take this approach: 1) Learn skills via courses 2) Demonstrate skills via projects 3) Apply for jobs People have got jobs using this approach, but with the job market more competitive than ever a new approach is needed. I've talked to graduates from bootcamps (who use this approach) and are struggling to get jobs. The reason they struggle is they only think about what jobs to apply for AFTER they've done their projects. This means they have a portfolio of random projects unrelated to the job (think titanic, Boston housing, movie sentiment analysis...). With hundreds of aspiring Data Scientists applying for the same jobs, it's no wander they don't stand out. Here's what you should be doing instead: 1) Research and think about the industry you want to work in (e.g., healthcare, project management, nutrition) 2) Build Tailored Projects 3) Apply to jobs in that industry Employers will immediately see you've taken care with your application and care about the problems faced by the business, instead of being another random Linkedin easy apply. That's what my mate Lekan is currently doing - he wants to utilise his experience in project management o get a Data Science job in the same field, so we're building a Tailored Project to predict which major construction projects (e.g., HS2) are likely to fail and need support. So don't listen to bootcamps - focus on the industry you want to work in and backsolve from there. And that's before I get started on all the unnecessary stuff they teach you... If you want to learn more about how to do this, you can learn more via my newsletter: https://lnkd.in/eTAmRsDu (P.S you'll also get a free guide on Tailored Projects/the lowdown on my love life ;) )

4

2 Comments

If you come from the math/natural sciences-side of data science and haven't received any formal training in software engineering, i.e. moi, pleeease check this book out. Just from the chapter on OOP and the section on using Docker, I've learned so much. Once I read this thoroughly, I know I'll be referencing this again and again.

8