Liubomyr Bregman

Hello network, Valuing convexity corrections requires evaluating quadratic payoffs. Traditionally these are valued by replication. At least, until the seminal papers of Patrick Hagan, Woodward, Andrew Lesniewski , and Skoufis where an analytical approximation is found. Inspired by them, David García Lorite and I have written a paper where we find an analytical approximation for quadratic payoffs using techniques similar to those used by Christian Kahl (HyP Hyp Hooray paper). Also, by using Clark-Ocone formula, we establish a model-free connection between the CMS price and the quadratic payoff price. We also derive an approximation for the RFR index case. All the approximations are tested numerically to evaluate their performance. The code written is in Python ready for use and play. Paper: https://lnkd.in/dXgNJtBT Enjoy it! #CMS #derivatives #Python #endofreplication #convexity #github

36

8 commentaires

Excited to share my latest entry-level project on GitHub! I've developed a Python script for assessing investment risk tolerance based on user input, perfect for those new to financial analysis and Python programming. This interactive tool guides users through questions to determine their risk profile, providing personalized investment suggestions including value stocks, growth stocks, bonds, and a 60/40 portfolio. Simply clone the repository and run the script locally or at https://lnkd.in/dtdG9z6m to explore and learn. Check it out here: https://lnkd.in/dwWTbdQM Looking forward to your feedback and contributions! Let's empower smarter investing decisions together. 💡 #Investment #Python #GitHub #Finance #Investing #OpenSource #LearnToCode

1

🚀 Polars Just Got Even Faster: After diving deep into "Polars vs Pandas" in his previous article(https://lnkd.in/dV5-_z3e), Lazar Gugleta is back with news for data enthusiasts: "Polars is now up to 13x faster" thanks to NVIDIA’s RAPIDS! (https://lnkd.in/dv9iPduA) 🎉. This speed boost is made possible through GPU acceleration with the cuDF library, integrated right into Polars' Lazy API. The best part? You still get Polars’ simple, familiar syntax—now with next-level performance. Want to try it out? 1️⃣ Install: `pip install -U polars[gpu] --extra-index-url=https://pypi.nvidia.com` 2️⃣ Run your query with `engine="gpu"` in `.collect()` to tap into GPU speed. 💥 If you’re working with large datasets, Polars on GPU can seriously cut down processing time, costs, and make workflows smoother. #Python #Polars #GPU #NVIDIA #RAPIDS

55

5 commentaires

Peter Gostev 🔥 Wow, what an intriguing battle of the giants! It's fascinating to see the Elo ratings for the top 5 players in the AI arena over the past year. The inclusion of fresh data for Google models adds even more depth to the analysis. Your approach of using Gemini 1.5 Pro for writing down the lyrics and Suno for generating the voiceover shows innovative thinking and dedication to the project. 🎤 Looking forward to seeing more updates on this exciting competition! #AI #EloRatings #Innovation 💡

1

The incorporation of multimodal models into AI agents promises to bring about a new era of highly interactive and perceptive agents that can seamlessly navigate and respond to the complexities of real-world environments. Read my new article on the future of Human-AI Interaction. #HumanAiInteraction #HAX #GenAi #MultimodalAi #HCI

14

1 commentaire

This is a great announcement 😃 #dataviz #datavisualisation #datavisualization #charts #graphs #data

51

6 commentaires

In my latest article, I explore the fascinating evolution of Large Language Models (LLMs). These models have transitioned from theoretical concepts to influential tools that significantly impact our daily routines. Dive into the article to learn more about how LLMs are playing a crucial role in bridging the realms of human cognition and artificial intelligence. #GenAi #Artificialintelligence #LLM #Technology #Evolution

39

2 commentaires

With GPT-4o now out, the lecture generation has become even better! https://lnkd.in/eWMAnvVA As with the previous posts, the video here is generated completely automatically requiring me to only specify the topic and number of slides. While GPT-4o is now a bit faster, during the creation process the code goes through multiple steps of improvement by self-evaluating the plots and contents of each slide making sure that the information is both correct and interesting. This means that each video still takes about an hour to generate. But due to the decreased costs of the GPT, the video now costs about $5 to make. While still on the high side, this is making it easier to now create a full curriculum using the framework (https://lnkd.in/e9-HVKVa) which will be trialed as part of the online Data Science Masters course at UVA.

16

AI’s Role in Sustainability and ESG: A Double-Edged Sword. AI has the potential to drive advancements in sustainability, offering new tools and insights. Meanwhile, the environmental impact of AI cannot be overlooked as shown by Microsoft’s investment in AI which has led to a substantial increase in the company’s carbon emissions, imperiling its climate goals. Read my take on the topic here: https://lnkd.in/dsMB3bFX

14

1 commentaire

AI news from around the world in the past week look more or less like this: In a groundbreaking achievement, AI has reached a silver medal standard in the International Mathematical Olympiad, solving 4 out of 6 complex math problems. Meta released Llama 3.1, a massive AI language model with 405 billion parameters and Mistral released a fresh new flagship model Large 2. OpenAI announced free fine-tuning for their GPT-4o mini model and the introduction of a new AI-driven search engine, SearchGPT. Elon Musk’s AI startup, xAI, has activated what it claims to be the world’s most powerful AI training cluster. https://lnkd.in/dsE2bYsk informs me, that they have recorded 46 (!!!) new AI tools during last 7 days. It’s enough to drive crazy. Or at least make it hard to keep up. Anything else coming to your mind? Feel free to share in comments. PS. Simultaneously, there is ongoing race in companies. Everyone is looking for or already implementing AI solutions, that will contribute to increased efficiency and automation. What is the perspective of the employees? The latest research published by Upwork sheds some light on this. Nearly half (47%) of workers using AI say they have no idea how to achieve the productivity gains their employers expect. Over three in four (77%) say AI tools have decreased their productivity and added to their workload in at least one way. Therefore, alongside implementing the appropriate technology, we should focus on trainings, communication and dialogue. AI will not yield results al long as it does not become a true companion for people in their daily work.

8

Feeling a bit tired of the "Dolce Far Niente" during my vacation, I started flipping through a little book on stochastic calculus. While looking at an exercise in the book, I had the following reflection: Indeed, the work of quants requires a lot of rigor and manipulation of mathematical objects, which generally requires verifying several hypotheses. However, when working as a quant and under some pressure from traders to quickly move things forward, one might sometimes skip these verifications, which can lead to completely off-track results and constitutes an operational risk. For example, the goal of this exercise was to calculate lim n->oo E[|Xn - K|] where (Xn)n>0 is a sequence of positive random variables with constant expectation equal to a and verifying lim n -> oo Xn = 0 a.s. Generally, this type of exercise is solved by interchanging the limit and the expectation, but it is still necessary to verify that this is possible, either by the Dominated Convergence Theorem or the Monotone Convergence Theorem. The somewhat hasty quant who says, "Anyway, dominated convergence is always satisfied, and we're no longer in grad school to waste time verifying it" might be severely mistaken in thinking that lim n->oo E[|Xn - K|] = K , which, in my opinion, could make them lose a lot of money! In fact, in this case, the hypotheses of the Dominated Convergence Theorem are not satisfied, and by rewriting |Xn - K| = Xn + K - 2min(Xn, K), it can be easily shown that lim n->oo E[|Xn - K|] = a + K! Bonus: The financial consequence of this result is that in a Black-Scholes model, the value of the straddle at the forward money of maturity T tends towards 2S0 when T tends towards infinity. I'll let you tell me why ;).

37

1 commentaire

Looks like Meta is more Open-AI than OpenAI! 😂 Meta released Llama 3.1 🦙 and the 405 billion parameter model outperforms GPT-4o on several benchmarks Not only that, but they also updated the license! You can fine-tune 🛠️, distil 🧪, and deploy it anywhere 🌍. This Llama 3.1 is the Linux(open-source) equivalent of AI! 🐧

29

Tic tac, tic tac… While many are relaxing this August or enjoyed the Olympics, the EU legislator is hard at work. On the 1st of August, they enforced the new AI Act legislation (which I still struggle to pronounce correctly). The AI industry and practitioners had been calling for regulations to prevent excesses, but the EU has far exceeded all expectations. This is now the strictest AI legislation in the world. Although it is said that we have two years to comply, this is only partially true. By February 2025 (just 6 months from now), AI systems deemed to pose unacceptable risks will already face prohibiltions. The AI Act is designed to be extraterritorial, meaning that even if you operate outside of the EU, you may still be subject to the AI Act if your AI system's output is intended for use within the Union. Given the lack of comprehensive AI legislation elsewhere, many companies will be using this regulation as their north star for ensuring responsible AI. The EU might just achieve its goal of setting the worldwide standard in AI regulation (Yes, EU can take the lead in some aspects of AI innovation too 😊). But let’s not panic. Most AI tools currently used by companies will fall into the High or even Limited/Minimal risk categories. This won’t block development, but MLOps processes will need to adapt, and teams will need education. Welcome to this brave new world—let’s get ready for it!

43

1 commentaire

New ways of computing are evolving every day, not only quantum or photonic. DNA Computing evolves fastly: new system stores data, plays chess, and solves sudoku puzzles. For a long time, scientists have considered using DNA as a computing device to complement traditional computers. As the volume of global data grows exponentially, silicon chips are struggling to keep up with the demands for storage and computation. Moreover, the rise of language models and other forms of artificial intelligence is increasing the need for alternative solutions. The issue with DNA storage is that it often gets destroyed when the stored information is "read." Last month, researchers from North Carolina State University and Johns Hopkins University found a solution. They successfully embedded DNA molecules, encoding multiple images, into a branched, gel-like structure resembling a brain cell. https://lnkd.in/diYcxVg4

42

6 commentaires