Oner Celepcikay, Ph.D.

Bonjour 👋, Data scientist and data engineer fellows, have you heard about #GenAI? #LLMs? #Transformers? It seems to be the next big thing in the data space and I've been receiving more and more inquiries about it. Unfortunately you won’t be able to train such models unless your company has $20M to spent on it. The good news is you don’t have to, to create powerful applications such as #RAG based ones. As a humble start, I will be sharing a couple of tips and sources to get started with this space and tools to get practical ideas and pathways for upskilling. Starting with the article from Google released in 2017, that opened all the great things we're seeing in this space. #OpenAI, #HuggingFace, #MistralAI, #Anthropic... I don’t wanna diverge here by talking about #Langchain (a must have library) My very first recommendation is to spend a month reading this article - "Attention is all you need." Make sure you deeply understand what a Transformer (the T in GPT) is and what the attention mechanism is. #generativeai #neuralnetwork #transformer #datascience

28

4 Comments

Join us for an enlightening workshop titled "New Approaches in Robotic Grasping: From Deep Learning, Reinforcement Learning to Graph Neural Network" at the upcoming ICCKE 2024 conference. In this workshop, we present the culmination of six years of intensive research conducted at the Human and Robot Interaction Laboratory, University of Tehran, about Robotics Grasping. I will present this workshop alongside two of my Ph.D. students, Hamed Ghasemi and Hamed Hosseini Ghasemi. Together, we will unravel the complexities of robot grasping from a deep learning perspective, shedding light on cutting-edge techniques, challenges, and future prospects. Date: 19th Nov, 13:00 Online https://lnkd.in/dqcGb8KE https://lnkd.in/dzXKb-ES Abstract: The workshop titled "New Approaches in Robotic Grasping: From Deep Learning, Reinforcement Learning to Graph Neural Networks" delves into the cutting-edge advancements in robotic grasp detection, a critical component for the automation of industrial tasks. Over the past six years, the Human-Robot Interaction Laboratory has focused on developing innovative solutions for grasp detection, utilizing a combination of geometric approaches and intelligent methods. This workshop will showcase the practical applications of these approaches through the development of a 3-DOF Delta robot and an advanced gripper. Attendees will gain insights into the integration of deep learning models for accurate and adaptive grasp detection, the application of reinforcement learning to enhance the decision-making capabilities of robotic systems, and the utilization of graph theory for optimizing grasp strategies and robot movements. In the workshop, we will also discuss how Transformers are revolutionizing robotic grasping and manipulation. These models, like GPT-4V and Gemini, enable complex robotic tasks with minimal training, using just input prompts. We'll review the latest developments in this area, including models like RT-1 and RT-2. By exploring these new approaches, participants will be equipped with the knowledge and tools necessary to advance the field of robotic grasping, contributing to the evolution of more autonomous and efficient robotic systems in industrial settings. Workshop Outlines: 1. Introduction to Robotic Grasping 2. Geometric Methods and Integration with Intelligent Methods 3. Reinforcement Learning (RL) Based Methods for Data Insufficiency 4. Graph Analysis and Modeling for Compatible Problems 5. Popular Simulators Overview 6. Practical Setup and Results Presentation 7. Applications of Grasping Over the Developed Gripper and Delta Robot 8. Latest Researches in Robotic Manipulation #robot #robotics #grasp #deep_learning #RL #reinforcement_learning

321

3 Comments

✅️ Missing Values | Preprocessing https://lnkd.in/gPSNE47m

2

Predictions based on Machine Learning and Statistics often rely on maximal likelihood, which can trap LLMs in biases inherited from the majority opinions of training data. This problem has been substantially addressed through conditional statistics, allowing for mitigation of biases and preventing the cancellation of minority perspectives. https://lnkd.in/gMXKqwyB

7

Now available on edX, CS50's new Introduction to Programming with R, a popular language for statistical computing and graphics in data science and other domains, with CS50's own Carter Zenke. Overview at https://lnkd.in/eY-Rw4hk. Enroll for free at https://cs50.edx.org/r. #education #community #r #statistics 

1,575

34 Comments

Often LLMs hallucinate because of semantic uncertainty due to missing factual training data. We propose a method to detect such uncertainties using only one generated output sequence. Super efficient method to detect hallucination in LLMs. See paper on ArXiv: https://lnkd.in/dS4Vxj7t

692

13 Comments

Imitation learning (IL) learns either the reward (or preference) model or the behavioral policy by observing the behaviors of an expert. Unlike other popular policy-based machine learning approaches such as reinforcement learning (RL), IL does not need to specify explicit reward signals, thus it can achieve generalization with very few observations of expert trajectory. In the #ICAPS2024 work of Qian Shao, Pradeep Varakantham, and Shih-Fen Cheng, we further extend the IL framework to consider not only rewards but also cost constraints. Our approaches are tested in open-source environments such as Safety Gym and MuJoCo, and demonstrate superior performance over the current state of the art. Our research will be presented by Qian Shao in Session 1 (Reinforcement Learning), 10am June 4, in KC 201. She will also be at the Poster Session: 16:30-18:00 June 4. Paper:   https://lnkd.in/gyGZdwkf arXiv: https://lnkd.in/gMjapFbq Code:    https://lnkd.in/gNgkmvHR #ICAPS2024 #ImitationLearning

49

New GAI Textbook Announcement (The textbook is being used by Stanford CS372: AI for precision medicine and psychological disorders) The past two years have been a significant period in the field of AI, during which we have observed substantial strides in generative AI technologies. Having dedicated over 5,000 hours to research and analysis, I have compiled this book to share a collection of findings and hypotheses regarding Large Language Models (LLMs)—detailing their operational mechanisms, and strategies to mitigate biases and reduce instances of hallucination. The central premise is to enhance context awareness and define human objectives with greater precision. By doing so, LLMs can more effectively mirror human linguistic behaviors, utilizing context-specific linguistic features to fulfill intended goals. SocraSynth: Socratic Synthesis with Multiple Large Language Models. https://lnkd.in/giC4474p

25

My Latest Project: Intelligent FAQ Chatbot Using Semantic Search in Python without Using any 3rd Party LLM API! With the increasing use of LLMs and Semantic Search in AI and ML, I’ve developed an intelligent chatbot that can be used with a Custom knowledge base and text data, designed to be used to answer frequently asked questions (FAQs) specifically on AI and ML basics. This chatbot leverages semantic search techniques, allowing it to respond accurately to user queries, and provide relevant and concise answers. Here's a deeper look into how it works: 🔍 How It Works: Embedding Creation: The chatbot transforms questions and answers into embeddings and numeric vectors that capture the semantic meaning of text. This makes it possible to determine the similarity between user queries and available answers. Semantic Matching: When a user asks a question, the chatbot compares the embedding of the query with embeddings of predefined answers, finding the closest match. Result Retrieval: Once the best match is found, the chatbot responds with the most relevant answer, improving the overall Q&A experience for users. 🛠️ Libraries Used: Transformers: Used for creating embeddings, enabling effective semantic search by capturing contextual meaning. NumPy: Facilitates handling and manipulating arrays for efficient computation of embeddings. Pandas: Helps manage data, allowing for smooth preprocessing of questions and answers. 💡 This project is an open-source initiative for anyone interested in exploring AI-based chatbots or applying semantic search to other domains. Check out the full project here on GitHub: https://lnkd.in/gu-w3YcY I'm eager to hear your thoughts and welcome any feedback or ideas on enhancing the chatbot! Salam Ullah Khan #AI #MachineLearning #Python #Chatbot #SemanticSearch #NLP #DataScience #OpenSource #ArtificialIntelligence #Transformers #FAQBot

9

1 Comment

For easy reference!!

2

📍Open AI paper on how large language models compare with smaller models in cases of compute-utilization efficiency. Useful in cases of allocation of compute in budget constraints. Equally as important as the "Attention is all you need" paper.

21

1 Comment

Happy for my dear students for their progress and their project. Starting with them from the basics of machine learning and mathematics until deep learning and how they can build a deep learning model then fine tuning LLMs. Thanks to the Ministry of Communications and Information Technology (MCIT), Egypt and Digital Egypt Pioneers Initiative - DEPI for letting me have the chance to teach this course and lead the students. Also, thanks to Hugging Face the platform that I really appreciate and go for many tasks and documentation for these lessons. #HuggingFace #DEPI #MachineLearning #DeepLearning

5

1 Comment

Happy to share our collaborative recent publication on "Delay probability in adaptive systems based on activation function of classical neural networks" in the Egyptian Informatics Journal (Q1, IF-5.0). Many thanks to all authors, specifically Prof. Vladimir Mladenovic, for all the efforts right from the conceptualization stage resulting in this outcome. About this contribution: A novel approach is introduced for discrete-time digital nonlinear systems, enabling the derivation of closed-form symbolic responses in the time domain as explicit functions of system parameters. This approach enhances understanding of system behaviour, the algorithm, and underlying processes, offering improved analytical insights. The paper also includes modifications to the existing algorithm to optimize its performance and applicability. This work simplifies the design and analysis of nonlinear systems, making these techniques accessible to engineers and researchers without extensive mathematical backgrounds. https://lnkd.in/g25Qg-AE

7

Subject: Call for Research Submissions: Cureus Journal of Computer Science (Springer Publisher) Dear Researchers, We are pleased to invite you to submit your research manuscripts to the Cureus Journal of Computer Science, published by Springer. This journal is committed to advancing the field of computer science by providing a platform for high-quality research in areas including cybersecurity, artificial intelligence, data science, machine learning, IoT, big data, and other cutting-edge domains. Important Note: All articles accepted for publication before the journal attains Scopus indexing will automatically be indexed once the journal achieves Scopus recognition. This ensures that your work will gain visibility and acknowledgment on a global platform. We welcome submissions from researchers, practitioners, and academicians. Our journal offers rigorous peer review and a dedicated editorial board, ensuring that all publications meet the highest standards of quality and impact. https://lnkd.in/dXrCJePb Prof.ammar almomani Associate Editor(Cureus Journal of Computer Science)

26

4 Comments

Pack just released a new book on the practical application of graph theory. It provides valuable explanations and practical examples of how graph-based algorithms implemented using Python solve industry problems, inspiring and motivating you to apply these techniques in your own work. It is an excellent book to expand your machine learning programming skills around network science. If you are interested in this book, please follow this link: https://lnkd.in/g2mD_CFY

30

5 Comments

Relase: 𝐒𝐢𝐧𝐚𝐓𝐨𝐨𝐥𝐬 - 𝐎𝐩𝐞𝐧 𝐒𝐨𝐮𝐫𝐜𝐞 𝐓𝐨𝐨𝐥𝐤𝐢𝐭 𝐟𝐨𝐫 𝐀𝐫𝐚𝐛𝐢𝐜 𝐍𝐋𝐏 𝐚𝐧𝐝 𝐍𝐋𝐔 يسعدنا إطلاق مجموعة أدوات ذكاء اصطناعي مفتوحة المصدر لفهم ومعالجة اللغة العربية Python APIs, command lines, online demos, and many datasets and language models. SinaTools outperforms all related tools in speed and accuracy, in all tasks: ▸ 𝗠𝗼𝗿𝗽𝗵𝗼𝗹𝗼𝗴𝘆 𝗧𝗮𝗴𝗴𝗲𝗿: Lemmatizer, POS tagger, root tagger. ▸ 𝗪𝗦𝗗 𝗧𝗮𝗴𝗴𝗲𝗿: Pipeline of semantic taggers: single-word WSD, multi-word WSD, and NER ▸ 𝗦𝘆𝗻𝗼𝗻𝘆𝗺𝘀 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗼𝗿: Extends a set of synonyms with more synonyms. ▸ 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰 𝗥𝗲𝗹𝗮𝘁𝗲𝗱𝗻𝗲𝘀𝘀: Association between two sentences across various dimensions, meaning, underlying concepts, domain-specificity, etc. ▸ 𝗡𝗮𝗺𝗲𝗱 𝗘𝗻𝘁𝗶𝘁𝘆 𝗥𝗲𝗰𝗼𝗴𝗻𝗶𝘁𝗶𝗼𝗻: Nested and flat NER, 21 entity types. ▸ 𝗥𝗲𝗹𝗮𝘁𝗶𝗼𝗻 𝗘𝘅𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻: Extract events and their arguments (agents, locations, and dates). ▸ 𝗗𝗶𝗮𝗰𝗿𝗶𝘁𝗶𝗰-𝗕𝗮𝘀𝗲𝗱 𝗠𝗮𝘁𝗰𝗵𝗶𝗻𝗴: Decides whether two Arabic words are the same taking into account diacratization compatibility. ▸ 𝗨𝘁𝗶𝗹𝗶𝘁𝗶𝗲𝘀: A set of useful NLP methods for sentence splitting, duplicate word removal, Arabic Jaccard similarity metrics, transliteration, and others. Try and Download: https://lnkd.in/dznzKb5b Birzeit University Birzeit University Alumni #arabic #arabicnlp #NLProc #arabicontology #LinguisticSearchIngine #lexical #Lexical_recourse #Sinalab #birzeit #Birzeit University | جامعة بيرزيت # #ArtificialIntelligence #ArtificialIntelligence #machinelearning #datascience

321

12 Comments

🚨ALERT: Innovation in artificial neural networks🚨 Published earlier today: “SWAG: A Novel Neural Network Architecture Leveraging Polynomial Activation Functions for Enhanced Deep Learning Efficiency” Kudos to Dr. Saeid Safaei, my last doctoral student at the The University of Georgia. This work is significant because it breaks a dogma in artificial neural network (ANN) literature. In 1993, Leshno et al. reported that an ANN “… can approximate any continuous function to any degree of accuracy if and only if the network's activation function is not a polynomial”. This myth became legend, the legend became dogma. We showed that not only it is possible to use polynomials… it is desirable. Our method with simple architectures outperforms traditional ANNs with complex architectures. It could have profound implications to decrease energy demand for AI through the development of photonic chips. https://lnkd.in/gQAtcNEa

23

8 Comments