Ayhan Diş

Python was invented in the late 1980s by Guido van Rossum at Centrum Wiskunde & Informatica (CWI) in the Netherlands as a successor to the ABC programming language, which was inspired by SETL capable of exception handling and interfacing with the Amoeba operating system.Its implementation began in December 1989. Van Rossum shouldered sole responsibility for the project, as the lead developer, until 12 July 2018, when he announced his "permanent vacation" from his responsibilities as Python's "benevolent dictator for life" (BDFL), a title the Python community bestowed upon him to reflect his long-term commitment as the project's chief decision-maker(he's since come out of retirement and is self-titled "BDFL-emeritus"). In January 2019, active Python core developers elected a five-member Steering Council to lead the project. Python 2.0 was released on 16 October 2000, with many major new features such as list comprehensions, cycle-detecting garbage collection, reference counting, and Unicode support.Python 3.0 was released on 3 December 2008, with many of its major features backported to Python 2.6.xand 2.7.x. Releases of Python 3 include the 2to3 utility, which automates the translation of Python 2 code to Python 3. Python 2.7's end-of-life was initially set for 2015, then postponed to 2020 out of concern that a large body of existing code could not easily be forward-ported to Python 3. No further security patches or other improvements will be released for it. Currently only 3.8 and later are supported (2023 security issues were fixed in e.g. 3.7.17, the final 3.7.x release. While Python 2.7 and older is officially unsupported, a different unofficial Python implementation, PyPy, continues to support Python 2, i.e. "2.7.18+" (plus 3.9 and 3.10), with the plus meaning (at least some) "backported security updates".#snsinstitutions #snsdesignthinkers #designthinking

1

Found this brilliant compilation of 30 papers on AI by Ilya Sutskever. - https://lnkd.in/eXPcDKsE Sharing it for you !! 💯

12

Since I had some spare time during the weekends I put something together: https://lnkd.in/eyjqrzbn This is the start of a new journey where I work on my passion of applying modern ML/NLP/AI/GNN techniques for Source Code (and Binary) Vunerability detection and Reverse Engineering. It features a new site, that I personally built, and where I planning to publish my personal Research and Tools. #NLP, #BERT, #CodeBERT, #GraphCodeBERT, #Transformers, #Cybersecurity

13

𝐂𝐚𝐧 𝐦𝐚𝐜𝐡𝐢𝐧𝐞 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐡𝐞𝐥𝐩 𝐮𝐬 𝐛𝐮𝐢𝐥𝐝 𝐛𝐞𝐭𝐭𝐞𝐫 𝐭𝐡𝐞𝐨𝐫𝐢𝐞𝐬? That’s the question we explore in our new PLOS ONE paper. 🔍 𝐓𝐡𝐞 𝐬𝐡𝐨𝐫𝐭 𝐚𝐧𝐬𝐰𝐞𝐫? Absolutely! Use machine learning for initial pattern discovery, then use those insights to shape the (causal) structure of your statistical model. A lot has been said about the superior predictive power of machine learning compared to traditional statistical methods. The common wisdom is that you have to choose: accuracy or explanatory insights — 𝘯𝘰𝘵 𝘣𝘰𝘵𝘩. This trade-off is false. In our latest PLOS ONE paper, we show that you can have both. We introduce “co-duction”—an approach that integrates machine learning within a structured five-step research process, combining the best of both worlds. 𝐖𝐡𝐚𝐭’𝐬 𝐍𝐞𝐰? While others have suggested using ML inductively, our approach is the first to integrate ML across a complete research process—incorporating inductive, deductive, and abductive steps. 𝐂𝐚𝐮𝐬𝐚𝐥 𝐀𝐈? 𝐒𝐨𝐫𝐭 𝐨𝐟! While most causal AI tries to identify cause-and-effect directly, our approach uses ML-driven pattern discovery to set the foundation for subsequent (causal) modelling. 𝐇𝐨𝐰 𝐭𝐨 𝐮𝐬𝐞 𝐭𝐡𝐢𝐬? We designed the paper to serve as a practical guide for scholars and data science practitioners alike. Skip to Table 1 (page 10) for the step-by-step! We’ve been using this approach successfully at The Big Data Company. By starting with ML-driven discovery, we capture potential patterns and validate them statistically—ensuring our models are theoretically sound and responsive to stakeholder needs. 𝐂𝐮𝐫𝐢𝐨𝐮𝐬 𝐭𝐨 𝐥𝐞𝐚𝐫𝐧 𝐦𝐨𝐫𝐞? Read it here: https://lnkd.in/gdtqGgfS Thanks to my co-authors Gwen Lee and Arjen van Witteloostuijn, and to all the colleagues and all (anonymous) reviewers who helped shape this work.

31

My colleague Jeroen Overschie made me finally understand how you can feed relevant documents into a large language model to enhance its context and provide better information and how important a good, relevant search engine still is for this method. In short: if your search engine and the results it turns up are better, the quality of answers from the LLM will be better.

9

An interesting post on how Skyscanner uses the combination of Dr Jekyll and its in-house statistical tool called WISE to ship experiment variants to production. Instead of using a frequentist approach as is done in most AB experiments, this framework uses a Bayesian approach to quantify the uncertainty in the final estimates by reporting credible intervals instead of confidence intervals. Summary: - Example of an experiment to test whether a pop-up, mentioning that ‘Skyscanner never takes a cut’ increases the ‘redirector rate’ or not. - If sufficient data has not yet been collected, WISE’s results will say ‘keep testing’. - WISE is able to provide a recommendation on the variant to ship. - WISE uses the Bayesian approach. They model their belief of the underlying parameter values as probability distributions, which are updated as new data is observed. - For conversion metrics, it uses a beta-binomial model - For revenue per user metrics, it uses a hurdle gamma-exponential model - What's the experiment stopping criteria: Through Monte Carlo simulations of thousands of experiments, they have determined a set of bespoke target 90% HDI widths to serve as their stopping criterion, which ensured an appropriate balance between the duration of experiments and the correct decision rate. https://lnkd.in/dv7BEHXe #statistics #analytics #rstats #python #peopleanalytics #experimentation

6

1 commentaar

🚀 Exciting Update: Just Completed DeepLearning.AI's Course on Large Multimodal Model (LMM) Prompting with Gemini! 🌟 I’m thrilled to share that I’ve completed an incredible course on Gemini, focusing on its advanced capabilities with images, videos, and real-time data using function calling: 🏞 1. Image Use Cases I explored how to build impactful tools like: Interior Design Assistants: Analyzing and recommending design elements from multiple images. Receipt Itemization: Extracting and organizing details from various receipt images. ▶️ 2. Video Innovations The course also covered: Semantic Video Search: Finding specific content in vast video datasets. Long-Form Video QA: Answering complex queries and summarizing lengthy videos. 🛠 3. Real-Time Data Integration I learned how to: Extend Gemini with APIs: Combine Gemini’s NLU with real-time data and interactive services via function calling. These insights have opened up exciting possibilities for developing advanced multimodal applications. A big thanks to DeepLearning.AI for this valuable learning experience! #AI #MachineLearning #DeepLearning #MultimodalAI #Gemini #Tech #Innovation #DeepLearningAI

15

6 commentaren

Two of my favorite things: machine learning and turbulence. Interesting read by Rik, Wouter, and Daan Crommelin. "Fully resolving turbulent flows is computationally expensive. Therefore, we typically work with under resolved simulations with added turbulence model. Recently, Neural Networks are being used as turbulence model with as target to reproduce the fully resolved flow field. This is a high dimensional problem. However, scientists are often only interested in a few quantities which are derived from these flow fields, such as the average energy. We propose a simple method which only uses these quantities as "target", resulting in a much lower dimensional problem. We compare this method with convolutional neural networks and find that both can successfully correct long-term simulations." Read here: https://lnkd.in/etBe9BnX Aside, if you are interested in this overlap of CFD/AI and want to get started in your learning journey -> checkout my book I have written on this topic: https://geni.us/JaXp #turbulence #ai #machinelearning #datascience #aerospace #aerodynamics #cfd #cae

95

1 commentaar

Closing thoughts on PyData Eindhoven, data scientists have a very good understanding regarding good uses cases for Generative AI, and, as well, use cases that don't really fit Generative AI and are better served by other approaches. For those use cases, the discussion is about monitoring, quality control and management of AI applications. My talk was the first talk in the back-up list, however, none of the titular roster speakers needed to excuse him/her self, so it would be next year. For those wondering what my talk was to be about, check the following repo :) https://lnkd.in/ehwC7h4H

12

Large Language Models (LLMs) are kinda dumb 🤔 This is quite a statement, I know. But I can explain how LLMs work by showing their strange behavior and why that happens. Meet 'tokens,' the building blocks of LLMs, similar to the alphabet in Latin languages. While we have 26 letters, LLMs like GPT-4 use around 100,000 ‘letters’ (tokens). When someone asks you to determine the number of times the letter ‘r’ appears in "strawberry", you can: ↳ Dissect the word into its letters ↳ You get: s-t-r-a-w-b-e-r-r-y ↳ Count each occurrence of ‘r’ ↳ The outcome will be 3 But if I asked you how many times the letter ‘r’ appears in the word ‘🍓’, what would your answer be? How would you dissect the word ‘🍓’? That’s pretty much what you’re asking an LLM to do when you ask it to count the characters in a word… Think about it. When you understand how LLMs work, operate, and think—with their 100,000-button keyboard—you can better grasp the tasks they can and can’t do. In the coming weeks, I’ll share more examples to help you understand how LLMs work and how to improve your way of working with them. Let’s connect if you want to know more.

60

6 commentaren

https://lnkd.in/dkaXy-ve Mathematicians and Engineers use ordinary differential equations to model and describe the behaviour of dynamic phenomena. In quantitative finance we use Stochastic Differential Equations ("SDEs") because asset prices are to great extent random. The most important Stochastic Differential Equation is called Geometric Brownian Motion ("GBM") introduced by the Great Economist Paul Samuelson in 1965. GBM assumes that asset prices follow a log normal distribution and that asset returns are normally distributed (a not so realistic assumption due to fat tails observed). Read the article below...

4

DocETL: Agentic Query Rewriting and Evaluation for Complex Document Processing Shankar et al.: https://lnkd.in/eHsJCY33 #ArtificialIntelligence #DeepLearning #MachineLearning

16

This latest video in Semih Salihoğlu graph database fundamentals series gives a truly excellent intro to RDF! It gently introduces RDF terminology along with concrete examples to help understand the benefits of #rdf and where it can be useful. Some Kùzu Inc. resources are also linked in the video description. Having studied these topics for a while myself, I remember how intimidated I was in my early days because it all felt so overwhelming, so I think this video will be very helpful for those who are getting started on their journey understanding the trade-offs between property graphs and RDF :). Looking forward to sharing part 2 of this video soon! #graph #database

5

in Pandas 2.2 one can set pyarrow as the default backend. in this presentation the advantages of arrow vs numpy are presented: 1. Faster string operations 2. Consistent missing value handling 3. Efficient data serialization Minute 24 https://lnkd.in/gbbGpR56

22

Battle of the Ensembles: Bagging vs. Boosting in Machine Learning In the realm of machine learning, ensemble methods like Bagging and Boosting have revolutionized predictive modeling. Both techniques aim to improve model performance by combining multiple models, but they do so in fundamentally different ways. ✨Bagging: Strength in Numbers Bagging, short for Bootstrap Aggregating, involves training multiple instances of the same model on different subsets of the training data. These subsets are created using bootstrapping—a method that randomly samples with replacement. Each model in the ensemble makes its prediction, and the final output is typically determined by averaging for regression tasks or voting for classification tasks. Bagging helps reduce variance and prevents overfitting, making it highly effective for unstable models like decision trees. A quintessential example of Bagging is the Random Forest algorithm. 📑Boosting: Building Stronger Learners Boosting takes a different approach by sequentially training models, where each new model aims to correct the errors of its predecessor. Initially, each data point is weighted equally, but as the process continues, the algorithm increases the weight of misclassified points, focusing more on difficult cases. This iterative method creates a strong composite model with significantly reduced bias and variance. Famous Boosting algorithms include AdaBoost, Gradient Boosting, and XGBoost. Boosting is particularly powerful for improving the performance of weak learners. 👩‍💻Key Differences and Use Cases - Error Handling: Bagging primarily reduces variance, making it ideal for high-variance, low-bias models. Boosting, on the other hand, reduces both bias and variance, which makes it suitable for a broader range of models. - Complexity and Training Time: Bagging is simpler and can be parallelized easily, leading to faster training times. Boosting, with its sequential nature, is more computationally intensive and can be harder to parallelize. - Performance: While both methods enhance performance, Boosting often outperforms Bagging in terms of accuracy but may be more prone to overfitting if not properly tuned. In conclusion, both Bagging and Boosting are powerful tools in a data scientist's arsenal, each with its own strengths and ideal use cases. The choice between them depends on the specific problem, model characteristics, and computational resources. For more in-depth learning and hands-on experience with these and other machine learning techniques, visit InfiniData Academy for comprehensive data science classes. https://lnkd.in/gcHK2QRP #machinelearning #datascience #baggingvsboosting #ensemblelearning #ai #randomforest #xgboost #InfiniDataAcademy #techtrends #learndatascience

5

🌟 Facebook Prophet for Time Series Forecasting 🌟 In this article, I'll walk you through the statistical overview of Facebook Prophet, shedding light on its inner workings📈 FBProphet is used to model time series data as a combination of Trend (T), Seasonality(S), and Noise Terms (E) by also incorporating random fluctuations or Holiday effects that might influence forecasts. It comes in handy if our data has multiple shifts in trend or has multiple layers of seasonality and can also handle non-stationary data. Even though the algorithm is based on a curve-fitting approach, it also incorporates the sequential nature of data. Prophet uses Hierarchical Bayesian Framework meaning it works on providing best-fitting posterior distribution to estimate coefficients of Trend and Seasonality equations, thus generating probabilistic forecasts that account for uncertainty around the point forecasts. In the comments, please feel free to provide feedback on any areas where more detail or clarification could be beneficial. Your insights will help me enhance my understanding and refine the article accordingly. In future posts, I aim to explain in detail the statistical processes and terms mentioned in the article. #FBProphet #TimeSeries

19

1 commentaar