Sebastian Raschka, PhD

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 

1,576

34 Comments

Shashank Sonkar and colleagues are back with another great paper and matching GitHub repo, "Pedagogical Alignment of Large Language Models". You can read the paper on arXiv here - https://lnkd.in/enZfCJgi, and browse the code here - https://lnkd.in/gyW5hSzp. They demonstrate an interesting approach to generating synthetic data to train open weights models to behave more pedagogically, as well as developing some new evaluation metrics. I think there's lots here for those of us interested in this kind of work to learn from and build on. (If synthetic data generation is new do you, check out this post from answer.ai earlier this week - https://lnkd.in/ecsG7R4b.)

38

4 Comments

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

The Problem of Peer Review https://lnkd.in/dsbQxmwv It’s been stated before, but an excellent paper in arXiv (“what can Natural Language Processing do for Peer Review?” spells out the problem of peer review very clearly, by describing what happens with papers submitted for a major conference: a vast amount of human effort. “For a concrete example, the ACL-2023 Conference alone generated over 12,000 reviews for over 4,500 full-text manuscripts … yet none of this data is public.” Clearly, when we think about open access, we should think about open reviews as well. NLP can’t fix everything connected with peer review, but this paper shows there is a lot that it could be doing right now. Incidentally, the authors, more than 20 of them, deserve praise for writing a wholly intelligible article (the outcome of a seminar), that doesn’t need a plain-text summary.

13

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

I thought it was very interesting to hear Andrej Karpathy talk at GPU MODE IRL 2024 about llm.c and the conversion of PyTorch code into C/CUDA, as well as the potential implications of this work for future LLM progress. The idea of converting high-level deep learning frameworks like PyTorch into low-level, highly efficient C/CUDA code could drastically improve model performance and execution speed, especially in specialized environments. It will be exciting to see how software engineering (SWE) practices evolve over the coming years as this becomes more widely adopted. One of the most intriguing aspects is the potential for a model to generate a specific kernel for a use case at someone's company in code that has no dependencies and could be fully optimized for that specific task. This capability could not only make models more efficient by reducing overhead but also streamline the workflow for developers, allowing them to focus on higher-level design and problem-solving. Essentially, it introduces a whole new paradigm for how we think about integrating AI into SWE roles. Instead of relying on pre-built libraries or packages, engineers might one day leverage AI models to dynamically create specialized, hyper-optimized code that meets their unique needs, reducing both the time and complexity of development. This shift could also lead to a greater level of automation, where AI models help developers by taking over tasks like performance optimization or low-level code generation, potentially reshaping the way we approach software development as a whole. You can listen to the keynote here: https://lnkd.in/eqJYS8Gb #AI #MachineLearning #DeepLearning #PyTorch #CUDA #SoftwareEngineering #C #LLM

2

Looking how to select better representatives from your clusters? Take a look at Lexin Chen's PRIME repo: https://lnkd.in/enM_Btbx Now with updated functionality and tutorials! Great collaboration with Arup Mondal and Alberto Perez! #moleculardynamics #clustering #machinelearning

65

𝗧𝗵𝗲 𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝗮𝗿𝘆 𝗱𝗲𝘃𝗲𝗹𝗼𝗽𝗺𝗲𝗻𝘁 𝗼𝗳 𝗣𝗘𝗙𝗧 To perform task-specific full fine-tuning with LLMs like LLaMA3 70B, a minimum of 5120GB of computational resources may be required. The enormous computational resource requirements are prohibitive for anyone but the superpower players to utilize LLMs for task-specific fine-tuning. To address this challenge, PEFT has emerged as a viable solution to compensate for the tremendous computational cost of full parameter fine-tuning. PEFT updates only a small number of additional parameters or updates a subset of the pre-trained parameters, preserving the knowledge and reducing the risk of catastrophic forgetting. As a result, there has been an increase on PEFT techniques during the last few months. Some of the most popular techniques where many customers are purring lot of effort to improve LLMs for theirs specific purposes are LoRA, QLoRA or Prompt-Tuning; If you want to start easily testing these techniques, I recommend to follow this couple of repositories: 👉 https://lnkd.in/dtvHP_Cn 👉 https://lnkd.in/daKDY4f5 #ai #genai #finetuning

72

3 Comments

🚫 Top 5 reasons why you should NOT apply for a Machine Learning / Artificial Intelligence position at TheraPanacea 🚫 Application Portal: https://lnkd.in/ePDdTgGv 1️⃣ You're just not into the latest trends in Artificial Intelligence 🤖... too futuristic for you, right? 2️⃣ Helping advance precision medicine and providing the best treatment to patients? Nah, you prefer things to stay... imprecise 🧑‍⚕️❌. 3️⃣ The thought of collaborating with the best clinical and academic centers worldwide doesn't give you any goosebumps 🌍💼... 4️⃣ You’re allergic to international, interdisciplinary, and gender-balanced teams 🌐👩‍💻👨‍💻. 5️⃣ Paris? ... who likes croissants, the Eiffel Tower, and vibrant culture anyway 🥐🇫🇷? But hey, if none of these sound like you, maybe you’re just the person we're looking for 😉! #AI #MachineLearning #Hiring #JoinUs

107

3 Comments

I tried to explain LLM hallucination intuitively using the concept of semantic entropy. I treated the two different scenarios of short response and long response separately. Hallucination mechanism with long response is more complex since the output typically weaves through multiple concepts. #ai #llm #hallucination #semantic #entropy https://lnkd.in/gF2_2H6w

35

5 Comments

🚀✨ Paper Alert! ✨🚀 Absolutely thrilled to share the publication of our latest paper entitled "𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐦𝐞𝐧𝐭 𝐨𝐟 𝐓𝐢𝐦𝐞-𝐃𝐞𝐩𝐞𝐧𝐝𝐞𝐧𝐭 𝐂𝐨𝐡𝐞𝐫𝐞𝐧𝐭 𝐁𝐚𝐜𝐤𝐬𝐜𝐚𝐭𝐭𝐞𝐫𝐢𝐧𝐠: 𝐋𝐞𝐯𝐞𝐫𝐚𝐠𝐢𝐧𝐠 𝐏𝐡𝐚𝐬𝐞 𝐈𝐧𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧" in Physical Review B. In this work, we show how the angular framework could be leveraged to characterize multiple scattering in a way that has never been done before. Instead of measuring the width of the incoherent cone as is done in SOTA techniques, we have an even better idea! We exploit the coherence of complex signals backscattered from different directions. We therefore enhance the coherent cone and we measure its width! This significantly facilitates the measurement of transport properties and paves the way to in vivo applications e.g., characterization of lungs or bones! Kudos to Baptiste HERIARD DUBREUIL for the amazing work!

32

1 Comment

I recently read “Towards Monosemanticity: Decomposing Language Models with Dictionary Learning” by Bricken et al 2023. Published at transformer-circuits.pub In this paper the authors discuss the use of autoencoders to extract interpretable features from a one-layer transformer model. The approach aims to improve the understanding and transparency of artificial neural networks. To do this they use the concept of dictionary learning which is a method used to decompose the activations of the transformer model into a set of basis features that are sparse and interpretable. This involves training sparse autoencoders to identify a dictionary of monosemantic features (or features with a single meaning). These monosemantic features in single units (artificial neurons) reminds me of the “grandmother neuron” theory in neuroscience which proposes that specific neurons respond to specific highly complex stimuli. However, in real brains, it is widely believed that complex representations are encoded by the collective activity of groups of neurons (and not monosemantic). But going back to the paper, I really like the visualization tools and illustration to make such complex networks more transparent (as opposed to black boxes). Paper: https://lnkd.in/gR87pJcs

3

Synthetic Data for LLM applications Synthetic healthcare datasets have been around for some time. Whenever we discuss synthetic data, we often find ourselves trapped in one question: How realistic is the synthetic data? We create synthetic data because we cannot share the original data for various obvious reasons, e.g., privacy, security, etc. The challenging part is that we want the synthetic data to have all the nice characteristics of the original data. With all the challenges aside, many researchers adopted and tried to prove the value of synthetic data. This morning, I came across this well-summarized survey article, "Synthetic data in healthcare: A narrative review," by Aldren Gonzales, Guruprabha Guruswamy, and Scott Smith. According to the paper, people have used synthetic data for seven categories: 1. Simulation and Prediction Research 2. Hypothesis, Methods, and Algorithm Testing 3. Epidemiological Study/Public Health Research 4. Health IT Development and Testing 5. Education and Training 6. Public Release of Datasets 7. Linking data I have been thinking about synthetic data quite a bit these days. It's because of the Large Language Models (LLMs) like ChatGPT by OpenAI and Claude by Anthropic. I have been playing around with those LLMs. I realize these days that, even if you are an expert, it's almost impossible to predict the outputs. As we cannot logically bind the output behaviors, testing an LLM model with all possible(or available) cases is the only way. Here, I think, is the new application of synthetic data: LLM application testing. If we were to build many LLM applications, we would need a lot of synthetic healthcare data. In this case, the synthetic data do not need to be hyper-realistic. They just need to represent the "quirkiness" of the real-life use cases, so weird synthetic data should be fine. Healthcare-focused LLM applications would need to test with all sorts of available synthetic data to see if such applications produce weird outputs. We may need a mechanism to do so. I think this new use case of synthetic data will be critical in healthcare. Let's see. [1] https://lnkd.in/e8DiEH9j

64

9 Comments

Anthropic has unveiled groundbreaking insights into the inner workings of Large Language Models (LLMs) in a recent paper. This pioneering research employs "dictionary learning" to map and manipulate the internal states of these models, offering unprecedented transparency and control. This breakthrough not only enhances our understanding of neural networks but also opens new avenues for ensuring the safety and reliability of AI systems. Read more: https://lnkd.in/eAzYgnKj IdentifAI - Find Origin - #artificialintelligence

236

3 Comments

The Information explores how OpenAI's reasoning models could help counteract potentially slowing progress from pretraining of larger more sophisticated models with more incremental gains over earlier models. It also notes that "relatively few developers" are using o1 models. While I haven't seen the full-fledged o1 models (and am certainly not a developer), the current models sometimes think hard about how to break code -- have seen them replace functional authentication with placeholders that you have to go back and update later, or swap out the correct variables with wrong ones -- for instance upper- or lowercasing something you didn't ask for, or adding an underscore where you didn't want one. At present, I think it still is better to have tighter feedback loops with the user -- otherwise, the genAI is thinking long and hard -- while a game of telephone can ensue where mistakes sometimes increase in the process of reflection -- about something that might not be what you wanted. A snippet from the paywalled article. The Information has done a lot of unique reporting on the subject.

4

1 Comment

Self Alignment using Instruction Backtranslation Recently, it's become common practice to fine-tune LLMs on proprietary datasets or domain-specific data. However, this process requires labeled data – often referred to as ground truth – and getting humans to label this data is both time-consuming and expensive. A recent paper, 'Self Alignment by Instruction Backtranslation',  where researchers have come up with an innovative approach to generate high-quality training data. Check it out here at: https://lnkd.in/dfx5yyha In simple terms, the process starts with a model fine-tuned on a small amount of seed data. The seed model constructs training examples by generating instruction prompts for web corpus. High-quality examples from among these candidates are selected and used to fine-tune a more aligned model. The cool part? When they tried this method with LLaMA, the resulting model outperformed all other LLaMA-based models on the Alpaca leaderboard not relying on data distillation. I've put together a code implementation in my GitHub repo. Feel free to check it out and let me know what you think! #GenAI #LLMs #Llama #Fine-tuning #MachineLearning #NLP #DataScience

47

Now online - very cool! Making Systems Research More Relevant: Lessons and Experiences with HarmonyOS NEXT https://lnkd.in/g6WK-VNK The relevance of systems research is often measured by its impact on industry practices. This talk revisits the lessons and experiences of conducting systems research crossing academia and industry by the speaker, who has spent around 8 years with various systems projects such as HarmonyOS NEXT and its open-source version OpenHarmony. It then uses HarmonyOS NEXT as a case to illustrate the potential of systems research to drive tangible innovations in the intersection of academia and industry. HarmonyOS NEXT explores many brave research ideas into practice, including microkernel-based architecture, distributed mobile computing and AI integration. By examining its architecture, features, and development process, this talk aim to identify key factors contributing to its success and to extract lessons for future collaborations between researchers and practitioners. By showcasing the potential of HarmonyOS NEXT, the speaker hopes to encourage further exploration of how academic insights can be translated into innovative approaches for complex systems challenges.

3

In case anyone is looking for an easy to use python package for working with complex valued functions or using arbitrary precision, mpmath (https://lnkd.in/eRGHrd2K) has some good things going for it! Demonstration of some capabilities below 👇 https://lnkd.in/eUHQBUW3

3

Can't believe I'd not put this paper on Research Gate or Google Scholar until today! Data Compression and Feature Selection: "Feature Selection for Supervised Learning and Compression" https://lnkd.in/e_USPuea My first paper from this work is still not able to be published! I finished it in 2019.

9

2 Comments