Cederik Haverbeke’s Post

🎙️ New Podcast Episode: The Rao-Cramer Inequality—Where Precision Meets Boundaries 🤯 What if mathematics could tell us the ultimate limit of precision in estimation? 📊 Our latest math-heavy podcast episode dives into the Rao-Cramer inequality, a foundational result in statistics that defines the lower bound for the variance of any unbiased estimator. 🔍 What You’ll Learn: 💠 Origins and Insights: Explore the deep connections between information theory and statistical estimation as established in C.R. Rao’s groundbreaking 1945 paper. 🥘 The Role of Fisher Information: Understand how this concept quantifies the information a dataset carries about its parameters and its impact on estimation precision. 🌏 Real-World Applications: From engineering signal processing to optimizing machine learning algorithms, discover how this inequality shapes decision-making and model design. 🌟 Mathematical Breakdown: We break down the inequality’s elegant formulation, explaining why it’s a benchmark for statistical performance. 💡 Whether you’re a data scientist, statistician, or math enthusiast, this episode will challenge you to think critically about what precision means in the context of uncertainty. 🚇 The episode is available on youtube and spotify. 🎤 Daniel A. Nir Regev #Statistics #MathPodcast #RaoCramerInequality #FisherInformation #DataScience

Our Latest episode is on the Cramer-Rao bound, specifically, Rao’s seminal paper.

Incredible https://lnkd.in/eurAZhgz While LLMs perform poorly on math problems, a combination of LLM and alpha proof agent may well be able to tackle some of the most complex scientific problems. All of this covered in a very lively conversation on Hardfork podcast, give it a listen! https://lnkd.in/e6U7W4Xe

It's unreal to be able to speak to folks like the guests I hosted on the SuperDataScience Podcast last month. ICYMI, today's episode highlights the most riveting moments from October. The specific conversation highlights included in today's episode are: 1. UC San Diego neuroscience professor Dr. Bradley Voytek on how data science facilitates breakthroughs in our understanding of the brain. 2. Eloquent Natalie Monbiot on how lifelike, digital versions of ourselves can scale up our public-facing work. 3. Lightning AI CTO Dr. Luca Antiga on where he sees generative A.I. being most useful in our professional lives. 4. Gable CEO Chad Sanderson on how, when we work with data, we always need to think about how downstream users might come to interpret our data... which is why he finds data contracts so important that he's writing an O'Reilly book about it. 5. Polars CEO Ritchie Vink on the incredible specs (e.g., efficiency speedups) of his open-source DataFrame-operations library for Python. Check out today's episode (#834) to hear all these eye-opening conversations. The "Super Data Science Podcast with Jon Krohn" is available on all major podcasting platforms and a video version is on YouTube. #superdatascience #datascience #machinelearning #ai #podcast

Interesting podcast. Tenure is definitely on the way out in many cases - primarily because it prevents (or at least forestalls) colleges and universities from making necessary programmatic changes to remain solvent. And although financial exigency is an option, it is tantamount to filing Chapter 11 - a public red flag. What will replace it? Likely 5- or 7-year contracts.

I recently tuned into an excellent episode of the Latent Space Podcast featuring Alistair Pullen from Cosine. The topic? Chain-based fine-tuning. This isn’t your typical domain-specific fine-tuning. It’s about teaching models to think, plan, and execute through structured reasoning chains. One standout example: Cosine fine-tuned GPT-4 to reason like a software engineer—not just on domain-specific data, but on structured thought processes. The innovation? Their data pipeline generates human-like reasoning traces, enabling models to tackle tasks with newfound sophistication. This approach aligns closely with OpenAI’s O1 reasoning model, which integrates reasoning chains directly into the model. It’s sparked some big questions for me: 🔹 How could reasoning-based fine-tuning reshape other domains? 🔹 Are we shifting from “what models know” to “how models think”? I’m eager to see how these methods evolve—especially with open-source alternatives in the mix.

Today I am shocked! I have probably only three times in my lifetime been stumped speechless. Today is the 4th. If you have not seen Google's NotebookLM (https://notebooklm.google/) it is an amazingly powerful interface that allows you to turn any information into accessible podcasts (as well as other things). Consider that I uploaded an article from 1998 on Quantum Computing and asked it to create a 10 minute podcast with interaction between two people. This is what it created: https://lnkd.in/endPiDZF in less than 5 minutes. The jokes, the interplay, the um's and uh's. I am stunned. Wow...can you imagine being a student and organizing material into a conversational delivery format for ease of understanding? The future for education is very bright. (Also, if you are interested in my actual podcasts I post one a week where I play Connect Four with amazing people as we talk for 4 minutes about topics in higher education. Note: not AI created!). Tagging my teaching and credential Techies and others who will be interested: Robert Bajor 🛰 , Allison Hall, Noah Geisel, Sheryl Grant, PhD, Ian Davidson, Kelly Page, PhD.Luke Hobson, EdDVinnie Rege, Ph.D.Meena Naik, Elizabeth Nowakowski,

In Episode #48, Yifei Yang introduces #predicate #transfer, a revolutionary method for optimizing #join #performance in databases. Predicate transfer builds on #Bloom #joins, extending its benefits to multi-table joins. Inspired by #Yannakakis's theoretical insights, predicate transfer leverages Bloom filters to achieve significant speed improvements. Yang's evaluation shows an average 3.3× performance boost over Bloom join on the #TPC-H benchmark, highlighting the potential of predicate transfer to revolutionize database query optimization. Join us as we explore the transformative impact of predicate transfer on database operations. Yifei Yang | Predicate Transfer: Efficient Pre-Filtering on Multi-Join Queries | #48 Disseminate is available on: - Spotify (https://lnkd.in/eHn-FaAF) - Apple Podcasts (https://lnkd.in/epQAjhYC) - Google Podcasts (https://lnkd.in/egTSSDva) - Amazon Music (https://lnkd.in/ehCt8Ypr) - Acast (https://lnkd.in/enfTYV-V) - Website (https://lnkd.in/efRkNaez) - Pocket Casts (https://pca.st/h7y2m0cx) Follow us on Twitter for more updates! https://lnkd.in/em4zVwpk You can support the podcast by Buying Me A Coffee: https://lnkd.in/e5AX9ABq #Research #ComputerScience #Data #Developer #Podcast #DataManagement #Databases #QueryProcessing #Joins #PredicateTransfer #MultiWayJoins #TPCH #Analytics

𝐇𝐨𝐰 𝐓𝐨 𝐁𝐫𝐢𝐝𝐠𝐞 𝐭𝐡𝐞 𝐆𝐚𝐩 𝐁𝐞𝐭𝐰𝐞𝐞𝐧 𝐓𝐡𝐞𝐨𝐫𝐲 & 𝐏𝐫𝐚𝐜𝐭𝐢𝐜𝐞 𝐰𝐢𝐭𝐡 𝐏𝐞𝐫𝐩𝐥𝐞𝐱𝐢𝐭𝐲 𝐀𝐈 Struggling to see how educational theories like Piaget or Vygotsky apply to your everyday practice? Connecting theory to practice can feel daunting at time no matter your level of experience. 𝐇𝐞𝐫𝐞'𝐬 𝐚 𝐬𝐢𝐦𝐩𝐥𝐞 𝐩𝐫𝐨𝐟𝐞𝐬𝐬𝐢𝐨𝐧𝐚𝐥 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐦𝐞𝐧𝐭 𝐡𝐚𝐜𝐤 𝐮𝐬𝐢𝐧𝐠 𝐀𝐈... ✨Ask Perplexity AI questions that link specific theories to your curriculum or framework. ✨Get practical examples and perspectives on how to implement theoretical concepts in your setting. ✨See how theories translate into actionable, realistic strategies for supporting children's learning and development. You can use Perplexity AI to explore how your favourite theory connects to your current planning - you'll find out how in this episode of the Early Education in the Age of AI podcast. I explain how Perplexity can become your go-to research assistant, helping you quickly and accurately gather information to better inform your overall planning and documentation and connect to frameworks like the Australian EYLF.  If you want to improve your understanding of educational theories and frameworks, making your job as an educator more efficient and less overwhelming then this episode is a must listen for you! 𝐋𝐢𝐬𝐭𝐞𝐧 𝐨𝐧 𝐒𝐩𝐨𝐭𝐢𝐟𝐲 𝐨𝐫 𝐨𝐯𝐞𝐫 𝐨𝐧 𝐭𝐡𝐞 E𝐦𝐩𝐨𝐰𝐞𝐫𝐞𝐝 E𝐝𝐮𝐜𝐚𝐭𝐨𝐫 𝐰𝐞𝐛𝐬𝐢𝐭𝐞 𝐡𝐞𝐫𝐞 >> https://bit.ly/41vyqPp