statistics.com

It's our final Elder Research Commercial Team data challenge of the year, and folks were given an image classification task: Is Santa in this image[1]? As usual, we saw lots of different approaches, too, from Resnets to vision transformers to LLMs. (DALL-E can generate some eerie-looking Santas, too.) The tools have improved a lot since I last looked at image classification! For this task, an off-the-shelf, zero-shot classifier (OpenAI's clip-vit-large-patch14 via HuggingFace[2]) performed super competitively, finishing just behind a coworker's fine-tuned Resnet 50 (F1 = 0.960 vs. 0.966). [There are lots of interesting subtleties in these results, too. For one, are these images actually new to the various pre-trained models, or were they a part of that pre-training? We also observe a nonzero level of corruption—or at least ambiguity—in the ground-truth labels, which brings down the maximum possible performance and adds some extra intrigue.] My first attempt at classifying with CLIP actually ended a little worse, with F1 = 0.85, and for an interesting reason. I had, effectively, asked the model to decide between two captions—Santa and not-Santa—and then made decisions based on a learned threshold for P(Santa). This was supposed to be a contrastive approach. But, plot the model's estimated log odds for the not-Santa query vs. its Santa counterpart, and it jumps out at you: CLIP can't tell much difference between Santa and not-Santa![3] The model thinks not-Santa is less likely on average in these data, but the log odds of not-Santa increase at almost the same rate as those of the Santa query; the model can't handle the negation (thanks, Robert Robison!) I did much better in a second attempt by dropping the not-Santa query completely, simply making decisions based on the log odds of Santa. So, things are pretty great at the moment for a simple case like this! More things are getting to be automatic, but these little surprises still get you from time to time; gotta pay attention. [1]: https://lnkd.in/eYjukzB7 [2]: https://lnkd.in/eZMNY7_7 [3]: Red diamonds are images labeled as Santa; blue points represent images labeled as not-Santa.

Sean Haler is back with a look at Advent of Code Day 2! He shares helpful tips for fellow data scientists, including: ✅ Using Boolean logic with Python’s any() and all() functions. ✅ Determining which sequences of data meet safety criteria. ✅ Tackling changing rules to evolve the solution 𝗡𝗼𝘁 𝗳𝗮𝗺𝗶𝗹𝗶𝗮𝗿 𝘄𝗶𝘁𝗵 𝗔𝗱𝘃𝗲𝗻𝘁 𝗼𝗳 𝗖𝗼𝗱𝗲? It’s an annual coding challenge that’s become a favorite among data professionals worldwide. Whether you’re a beginner looking to sharpen your programming skills or a seasoned pro eager to tackle clever puzzles, Advent of Code offers something for everyone. You can solve the puzzles at your own pace. So even if you just have 15 minutes each day, it’s a great opportunity to sharpen your skills. 💡 Advent of Code is one of our team’s favorite holiday traditions, and we’re proud to sponsor this year’s event! 𝗣.𝗦. Solving puzzles like these is what we love to do every day. Think you’d enjoy it too? Explore career opportunities with us: https://lnkd.in/e3eZxrQb #AdventofCode #DataScience #DataAnalytics #Programming

Hey data nerds! How are you sharpening your data science skills this season? 💡 In this video, Sean Haler, one of our principal data scientists, shares one of his favorite ways to stay sharp: Advent of Code! 🧑💻 Sean walks through Day 1 of this year’s challenge, using concepts every data scientist should know, including Python programming basics, the LP norm, and a weighted sum. Advent of Code started December 1, but it’s not too late to join the fun! You can solve puzzles at your own pace—even if you just have a few minutes a day to work on one. 🧩 And you can tap into the global Advent of Code community for tips and camaraderie. Our team loves solving the puzzles every year because it’s a fun way to sharpen our skills and learn together. We’re proud to sponsor this year’s Advent of Code. Hope you’ll join us and see why it’s one of our favorite holiday traditions! ❄️ Interested in joining our team? Check out our opportunities here on LinkedIn: https://lnkd.in/gpPJUkTT

Calling all coders!! 🗣️ There's still time to join in the fun! 🧑🏽💻 #adventofcode #statistics

We've been a bit quiet on LinkedIn lately... but for good reason! For the past few weeks, we've been working behind the scenes on something groundbreaking, and now it's finally here! 🌟 We’re excited to announce the launch of our brand-new Generative AI Foundations course! 🚀 Generative AI has grown so rapidly in the past year and many people are wondering, “what’s it all about?”. According to 451 Research's Generative AI Market Monitor & Forecast, they are projecting a compound annual growth rate of 59% through 2028 and observe that the generative AI market appears to be growing at an explosive rate. You WANT to be a part of this growth! 📈 Whether you're just starting out or ready to elevate your expertise, this is your chance to dive into the world of Generative AI and gain hands-on experience with the tools and techniques driving the future of innovation.   💡 Ready to explore the future? Don't miss the chance to grow your AI journey with us! ✨Just for our LinkedIn community, use the code 'GENAI LINK' for a course discount.✨ (see the link to register in comments) #GenerativeAI #GenAI #LifelongLearning #AIEducation #FutureOfWork #statistics #datascience

📈 LinkedIn's Meteoric Rise: The Power of Data Science and Algorithms LinkedIn has achieved an incredible milestone, becoming a $15 billion business in 2023! 🚀 But have you ever wondered what drives this impressive growth? The answer lies in the power of data science and sophisticated algorithms. Over the years, LinkedIn's revenue has skyrocketed, from $2.3 billion in 2017 to a staggering $15.1 billion in 2023. This type of growth isn't just luck – it's the result of leveraging data science to enhance user experience, optimize content delivery, and drive targeted advertising. 🔍 By analyzing user behavior, preferences, and interactions, LinkedIn's algorithms can personalize feeds, recommend connections, and suggest relevant job opportunities. This creates a more engaging platform, which in turn attracts more users and advertisers. Most recently, LinkedIn has added several cognitive-based games to further attract daily interaction on the platform. They’ve also introduced AI assistants to help users optimize their posts and draft articles. We’ve yet to see if these gamification and tech tools add-ons will indeed improve user experience. 🎮⚙️ Data science isn't just about crunching numbers; it's about turning data into actionable insights that drive business success. Discover how embracing data science can propel your business to new heights! #datascience #dataanalytics #algorithims #bigdata #predictiveanalytics #statistics

We are so excited to welcome Matt Bezdek to the Statistics.com faculty! 🎉🎓 Matt Bezdek is a Senior Data Scientist at Elder Research. He has over 10 years of experience in performing advanced statistical analyses. At Elder Research he helps commercial and nonprofit clients with model validation, software development, interactive data visualization, data literacy education, and building data analytic strategies. He holds a PhD in Cognitive Psychology from Stony Brook University and has conducted research at the Georgia Institute of Technology and Washington University in St. Louis. Matt will teach our Regression and Python for Analytics courses. What do YOU know about the topics of Regression and Python? 🤔

Good advice!✔️😊 When data scientist, Paige Spell first began in the field, she faced a few obstacles. 🪨 “Starting out as a data scientist, I was nervous to ask for help. I feared my questions would undermine my expertise,” says Paige. “Coming out of my master’s program, I was eager to apply every technique I learned, often overlooking the power of simplicity. It’s easy to get caught up in the day-to-day work, but this field changes so fast and you have to be able to change with it.”   From Paige’s perspective there are a few common mistakes new data scientists can experience. Here are her tips to help you avoid each of them:

"Generative AI is revolutionizing education, but how much do you really know about it? Whether you see it as a game-changer 😊 or a challenge 😣, understanding it is key! Take our quick quiz to find out where you stand! Click here: https://lnkd.in/eyw-XRqA *The first 10 quizzers get a prize!* Are you team 'good' 😊 or team 'bad' 😣? Share your results in the comments! #GenerativeAI #GenAI #AIQuiz #Education #EdTech #OnlineLearning #datascience

Quiz Time! 📊 How Much Do You Know About Generative AI? 🤔 Generative AI is a hot topic in education, sparking debates on whether it’s 'good' 😊 or 'bad' 😣. Whether you're optimistic about its potential or cautious about its implications, the first step to forming a well-rounded opinion is through education. 🧠 How much do you really know about generative AI? It’s time to put your knowledge to the test! Take our quick quiz to see where you stand and learn more about this transformative technology. Click here to get started: https://lnkd.in/eyw-XRqA So, which team are you on? Are you team 'good' 😊 or team 'bad' 😣? Let us know in the comments! #GenerativeAI #GenAI #Education #AIinEducation #EdTech #OnlineLearning #QuizTime #AIKnowledge #datascience