Lourens Walters

𝗙𝗿𝗼𝗺 𝗣𝗿𝗶𝘃𝗮𝗰𝘆 𝘁𝗼 𝗘𝘅𝗽𝗹𝗮𝗶𝗻𝗮𝗯𝗶𝗹𝗶𝘁𝘆: 𝗧𝗮𝗰𝗸𝗹𝗶𝗻𝗴 𝗔𝗜 𝗖𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝗮𝘁 𝗡𝗛𝗦 𝗘𝗻𝗴𝗹𝗮𝗻𝗱 I’ve recently written two articles for the NHS England Data Science Website based on some of the really interesting work I’ve been doing! The first: Investigating Annotation Tools for Named Entity Recognition (1). If you like Generative AI, entity extraction and identifying privacy risks - this is for you. The second: The RISE Tool (2) - An Easy Way for Testers and Assurers to Evaluate AI Classifiers. This covers AI evaluation, assurance, image classifiers and explainability. The NHS England Data Science team have a whole host of interesting projects, you can read about them all on our website (3). 1 - https://lnkd.in/dNaUXsgv 2 - https://lnkd.in/dEjp88ra 3 - https://lnkd.in/dYg53A4j

19

As Data Scientist we often get carried away with datasets, databases and even spreadsheets (in small Businesses). That we end up never finding the solution to the problem despite so much Trial and Error with ML, DL and even dashboarding. This is where the skill of Analogue Analytics comes in—a critical yet often overlooked aspect of our profession. What is Analogue Analytics? It’s the ability to approach problems with a human-first mindset, focusing on: ● Understanding the context: Why does this problem exist? What does "success" look like? ● Visualizing the data we should have: What might exist, and how does it connect to the problem? ● Recognizing nuances: Are there biases, trade-offs, or limitations in the available data? ● Thinking beyond algorithms: What patterns, behaviors, or trends could emerge from this situation? Why Does It Matter? ● Not every problem is a nail, and not every solution is a hammer. Without this analogue layer of thinking: ● We risk solving the wrong problems. ●Insights become detached from real-world impact. How Can We Develop This Skill? ● Ask better questions. Start every project with why, what, and who. ● Engage with domain experts. Their insights often fill the gaps where data falls short. ● Focus on storytelling. Simplify complex problems and solutions into narratives everyone can relate to. So what I want to say is that Data is powerful, but only when paired with the right perspective. Let’s not lose sight of the human element in data science. Because at its core, our job is to solve problems—not just analyze numbers. #datascience #psychology #management #humanmind #criticalthinking #softskills

27

1 Comment

Here's round up of this years tips and general advice for career in data science from Glasseye https://lnkd.in/eps5qxWu

3

Really happy to share that our work on “Learning Metrics that Maximise Power for Accelerated A/B-tests” will appear in the Applied Data Science track of #KDD2024. We address a common problem in industry: how can we minimise type-II errors in online controlled experiments? Can we simply learn a metric combination that directly minimises this quantity? This project started out as “let’s implement the ideas from this great paper by Eugene Kharitonov et al.”, but we ended up proposing some extensions to their seminal work that seem to be quite effective, and are now used across ShareChat. Proud to have this work out there, let us know what you think. Additionally, we were assigned an Area Chair that clearly cared: initial low-confidence reviews prompted them to invite more reviewers (we got 8!), and the AC themselves correctly grilled our rebuttals to point out arguments they deemed insufficient. It’s easy to be positive about peer review when your paper gets accepted, but it seems there are still some unsung heroes out there... Work with Aleksei U.. ACM SIGKDD & Annual KDD Conference

116

9 Comments

🌟 Retrieval-Augmented Generation (RAG) is truly a game-changer in the AI landscape! By seamlessly combining retrieval-based and generative models, RAG architecture enhances both the accuracy and relevance of AI outputs. The detailed process involving chunking, embedding, and vector search ensures that responses are not only precise but also contextually rich. This innovative approach is particularly beneficial in providing enhanced relevance, improved accuracy, and versatility across various domains. RAG is setting a new standard for intelligent systems, making AI interactions more informed and reliable. #AI #MachineLearning #RAG #ArtificialIntelligence #TechInnovation #DataScience #VectorSearch #LLM #Embeddings #GenAI 🚀🤖📚

2

This morning it was released my latest article in Towards Data Science, where I go over an end-to-end data science classification project following the #CRISPDM framework for data mining lifecycle. The project brought me good new learnings and brushed up some old skills: - Problem definition - Data exploration - Modeling - Dealing with unbalanced data The article is very complete and may become a mini-course on Udemy soon. Check it out, here: https://lnkd.in/ey6HbUmM #datascience #crispdmlifecycle #python #EDA #datamodeling

14

3 Comments

For me, this sentence captures the essence of the data-driven mindset. Interestingly, it wasn't said by Déscartes, Gauss or anyone we would usually associate with scientific thinking. It was written by Oliver Cromwell, on a letter to the Church of Scotland, advising them not to confront him, no matter how confident they felt that they would win. In Statistics, this became known as Cromwell's rule: we never associate any probability to be exactly 0 or 1, except for something that's logically impossible. #statistics #probability #datascience

11

2 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

I'm sure you know the data model for Kimball data warehouse, i.e. Kimball star schema, which is fact and dimension tables. That is called a dimensional data warehouse. It is a denormalised data model. But do you know the data model for Inmon data warehouse, i.e. the so called normalised data warehouse? How do we store the historical data with a normalised data warehouse. Does it have to be in third normal form? I was asking that question myself today, and found an old article that I wrote 14 years ago: https://lnkd.in/ezv5_imx It's about a Normalised Data Warehouse. I know the technology has moved on, new approaches like DV and LakeHouse are now popular, and a Normalised Data Warehouse is probably obsolete. But have a read and let me know your thoughts on the comments below. It's an old article though so apologies if it's not up-to-date and inaccurate/incorrect. It's 14 years ago. PS. Operational Data Store (ODS) is a normalised data store but it is not a data warehouse as it does not store historical data (only current values). But ODS has its own place in data warehousing though: it can be used as the middle/silver layer. #DataWarehouse #DataModel

51

10 Comments

Extended Long Short-Term Memory (xLSTM) was recently introduced by the authors of LSTM with the motive to see how far it can get in language modeling by scaling it up to billions of parameters. To overcome limitations of LSTM in storage capacity and parallelizability, the following modifications were implemented: 1. Exponential Gating : Replacing sigmoid gating with exponential activation functions for input and forget gates, offering broader dynamic ranges of gating responses. Supported by a Normalizer and Stabilizer to manage gate strength and prevent overflow from large values. 2. Enhanced Memory Structures: - sLSTM: Enables memory mixing within each head of recurrent connections, similar to the original LSTM, but not across multiple heads. - mLSTM: Upgrades memory cells from scalar to matrix for retrieval via matrix multiplication. Multiple heads and cells are interchangeable, unlike in sLSTM. Utilizes covariance matrices, similar to self-attention (Transformers), to capture evolving relationships over time for key-value pairs. Architecture-wise, residual blocks of the above mentioned memory structures are stacked, enabling the learning of complex relationships and patterns in the data through various transformations. - sLSTM Block with Post Up-projection (Like Transformers) - mLSTM Block with Pre Up-projection (Like State Space Models) In experiments, xLSTM architecture predominantly utilized mLSTM:sLSTM ratios of 1:0 and 7:1, showing superior performance. The paper suggests that xLSTM achieves comparable performance to existing architectures employing Transformers and State Space Models (SSM), with enhancements attributed to exponential gating and matrix memory integrations. https://lnkd.in/gXU3VCKS

51

#Closeread has the the potential to be a game changer in how we communicate about data and insights via a narrative. I have been captivated by Scrollytelling since the NYT times showed the spread of COVID-19 back in March 2020. https://lnkd.in/dGYwpR95 This tool is reducing the barrier to entry to create compelling narratives and I cannot wait to experiment with it

5

1 Comment

I’ve been thinking and talking about data visualisation a lot lately, so in that vein here is an outline of the differences between clustered bar charts and dumbbell plots (https://lnkd.in/gVdhjCX5), and some thoughts on creating fuzzy icon arrays (https://lnkd.in/ge3E3T-G; although I personally would steer clear of the photo grid options). Elsewhere, here’s a cool convenience function to speed up exploratory data analysis (https://lnkd.in/gFs672he), and a collection of writings on the human component of qualitative research (https://lnkd.in/gduTqWHC). Close to home for me is this article about efforts to create evidence banks for social science (https://lnkd.in/gGWufcC7), and I’m also charmed by these tiny translucent frogs (https://lnkd.in/g6kGWmEh). In good news, Indigenous rangers have found some of Australia’s most elusive birds (https://lnkd.in/g7gQuDc3), and the identification of the gene for plant thorns could lead to increased availability of Australian native foods (https://lnkd.in/gKfvnkWv). I really enjoyed reading this interview with a photographer who has a new collection of portraits of philosophers (https://lnkd.in/ggJn9tuX), and on the topic of photography here’s some incredible black and white shots (https://lnkd.in/gc_JmDFC). Those of you with an interest in technology might enjoy these ‘exploded’ artefacts (https://lnkd.in/gWiV3baM), while those of you with sesquipedalian tendencies may wish to augment your idiolect with some regional Australian slang (https://lnkd.in/gks-SQWK).

2

👉🏼 Reliability of large language models for advanced head and neck malignancies management: a comparison between ChatGPT 4 and Gemini Advanced 🤓 Andrea Lorenzi 👇🏻 https://lnkd.in/eSW4Admy 🔍 Focus on data insights: - ChatGPT 4 generally outperformed Gemini Advanced in adherence to guidelines and comprehensive treatment planning. - Higher AIPI scores were observed with ChatGPT 4, indicating better overall performance. 💡 Main outcomes and implications: - Both large language models showed potential in aiding multidisciplinary management of head and neck oncology. - Discrepancies in critical areas emphasize the need for further refinement and validation against current clinical standards. 📚 Field significance: - Growing role of AI in enhancing clinical decision-making. - Necessity for continuous updates and validation to fully integrate AI into healthcare practices. 🗄️: [#largeLanguageModels #headAndNeckMalignancies #AIclinicalDecisionMaking]

1

1 Comment

👉🏼 Expansive data, extensive model: Investigating discussion topics around LLM through unsupervised machine learning in academic papers and news 🤓 Hae Sun Jung 👇🏻 https://lnkd.in/ef_MF54v 🔍 Focus on data insights: - Utilized data from Web of Science and LexisNexis to explore topic modeling methods for Large Language Models (LLMs). - Evaluated various topic modeling approaches such as LDA, NMF, CTM, and BERTopic based on diversity and coherence metrics. - BERTopic demonstrated superior performance in diversity and coherence across both platforms. 💡 Main outcomes and implications: - News articles covered a wide range of topics related to LLM applications, focusing on specialized domains. - Research papers concentrated on technical aspects of LLM technology. - Insights gained can help understand future challenges and opportunities for LLMs in different sectors. 📚 Field significance: - Enhances understanding of how LLMs are utilized in academic and news contexts. - Provides insights for enterprises leveraging LLMs for service delivery. 🗄️: #topicmodeling #unsupervisedlearning #machinelearning #datainsights

1

Excited to share my latest article on Towards Data Science 📊 I reflect on eight years in data and share key lessons for aspiring data professionals. Check it out and let me know your thoughts!

32