Khan Siddiqui, MD’s Post

Excited to announce that our paper, "Progressive Inference: Explaining Decoder-Only Sequence Classification Models Using Intermediate Predictions," has been accepted to #ICML2024! In this work, we introduce a novel explainable AI (XAI) framework called "Progressive Inference". Our approach leverages intermediate predictions from decoder-only Transformer models to generate high-quality SHAP-like input attributions for sequence classification tasks. Our method significantly outperforms prior XAI techniques and is a step towards improving the trustworthiness of large language models. For more details, check out our arXiv preprint: https://lnkd.in/gC3z3t2J Joint work with Freddy Lecue, Saumitra Mishra, PhD, Chris Pond, Daniele Magazzeni and Manuela Veloso. #AI #MachineLearning #XAI #ICML2024

🚀 Smarter Transportation is Here! 🌍 We’re happy to share our work on using Large Language Models (LLMs) to improve transportation. Our study shows how LLMs can work with time-series, audio, and video data to make quick and smart decisions. 📊 What We Achieved: ✅ Built a system that works with different kinds of data. ✅ Reached 91.33% accuracy, doing best with time-series data. ✅ Made processes easier by using one model instead of many. 🚗 This system helps with tasks like planning and keeping vehicles in good shape, making transportation faster and safer. This was a team effort with my great co-authors: Dexter Le, Aybars Yunusoğlu, Karn Tiwari, and Murat Can Işık. Thanks to them for their hard work! Check out our paper here: https://lnkd.in/dAqvd_MH What are your thoughts on using AI for smarter transportation? Let us know! #AI #Transportation #Innovation #Research

Great step towards #trustworthy #LLMs, which are crucial for LLMs adoption at scale, specially when critical decisions are required to be made (#Finance domains…). Plenty of direct applications e.g., LLMs inputs attribution for #hallucinations detection, #FactChecker !! Even broader: think LLMs outputs now directly attributing LLMs inputs. Great work, Sanjay Kariyappa and team. LLMs inputs and outputs nicely connected with #XAI. #TrustworthyAI

The Advent of a New Epoch in GenAI? The OutEffHop Models Outperforming Conventional Attention Mechanisms in Transformers A revolutionary shift in AI is underway with the introduction of Outlier-Efficient Modern Hopfield (OutEffHop) models, which are challenging the traditional attention mechanisms in transformer architectures. Traditional attention has been fundamental in AI for tasks like language processing, but it falters when dealing with outliers. OutEffHop models leverage theoretical physics to improve data processing, treating outliers as an energy minimization problem, thereby increasing efficiency and reducing computational demands. This novel approach aligns AI more closely with the fundamental laws of physics, potentially leading to better real-world understanding in AI systems. The OutEffHop's adept handling of outliers enhances the signal-to-noise ratio, providing more accurate and reliable data insights. As these models gain traction, they could diminish the importance of traditional attention in large transformer models, marking a significant leap forward in AI capabilities. Businesses and researchers are prompted to investigate and embrace OutEffHop to maintain a competitive edge in technological innovation. The OutEffHop architecture heralds a smarter and more sophisticated generation of AI systems, characterized by intuition, efficiency, and robustness. For those interested in the technical details and potential impact of OutEffHop models, the full paper is linked in the post and warrants attention for its significance in advancing AI research and applications. #ai #llm #genAI #machinelearning #generativeAI #artificialintelligence #tech https://lnkd.in/evfd7pGa

Our team will present our latest graph learning work, "VCR-Graphormer: A Mini-batch Graph Transformer via Virtual Connections," at ICLR 2024 in Vienna, Austria. The paper introduces a novel approach for mini-batch based graph transformers, enabling them to efficiently process billion-scale industry graph datasets. This advancement opens new possibilities for research into large graph foundation models and has potential applications in fields like social network analysis and graph-based recommendation systems. We eagerly anticipate engaging discussions at ICLR 2024. Additionally, we invite you to visit us at the AI at Meta booth to explore more exciting AI advancements from Meta. @Dongqi Fu, Zhigang Hua, Yan Xie, Jin Fang, Si Zhang, Kaan Sancak, Hao Wu, Andrey Malevich, Jingrui He

Can You Turn Your Vision-Language Model from a Zero-Shot Model to Any-Shot Generalist? Meet LIxP, the Context-Aware Multimodal Framework Contrastive language-image pretraining has emerged as a promising approach in artificial intelligence, enabling dual vision and text encoders to align modalities while maintaining dissimilarity between unrelated embeddings. This innovative technique has produced models with remarkable zero-shot transfer capabilities, demonstrating significant potential in complex computational tasks. Read the full article: https://lnkd.in/duBaZr33 Paper: https://lnkd.in/d4KXs8fn

It’s day 4 of our 11 Days of Inference Acceleration Techniques. Today, we’re moving on to runtime level optimization best practices. 𝐅𝐨𝐮𝐫𝐭𝐡 𝐭𝐢𝐩: 𝐓𝐚𝐤𝐞 𝐚𝐝𝐯𝐚𝐧𝐭𝐚𝐠𝐞 𝐨𝐟 𝐠𝐫𝐚𝐩𝐡 𝐜𝐨𝐦𝐩𝐢𝐥𝐚𝐭𝐢𝐨𝐧 📈 Graph compilers such as TVM, Tensor-RT, and OpenVino work by getting a computation graph of a specific model and generating an optimized code adjusted for the target hardware. Graph compilation can optimize the graph structure by merging redundant operations, performing kernel auto-tuning, enhancing memory reuse, preventing cache misses, and more. But few things to be aware of: 📝 Not all models compile equally. 📝 The impact of compilation on model performance can vary. Hence, make sure to check the architecture's ability to compile early on in the process to avoid wasting time and resources on training a model that can’t be optimized for fast inference. – What’s the #11DaysofInferenceAccelerationTechniques? The Deci team is posting, for 11 days, a series of inference acceleration techniques for deep learning applications. If you’re looking for practical tips and best practices for improving inference, follow Deci AI so you won’t miss an update. #deeplearning #machinelearning #neuralnetworks #computervision

🔍 Zero-Shot Object Detection (ZSOD): Revolutionizing Computer Vision! ZSOD is transforming the field by enabling models to identify unseen objects, but it also has its challenges. In my latest blog, I discuss innovative approaches to overcome these limitations and how we used GPT-4 Vision to enhance object detection. Check it out for a deep dive into this cutting-edge technology! #genai #computervision #ZeroShotObjectDetection

Unified model architecture is a flexible framework that handles multiple tasks like vision, language, and audio using a single neural network. Unlike specialized models, it uses shared architecture and abstract representations for all tasks. Best practices involve training on diverse datasets and using sequence-to-sequence formats. Quality data labeling is important for optimal performance. Learn more about unified models and Sapien's data labeling for AI models: https://lnkd.in/gpJmMi-W #MachineLearning #AIResearch #DataScience

Considering the price ratio (1 : 14.3) and the TDP ratio (estimated in 1 : 20) for M1 max vs workstations with H100, these results seem promising and make me believe I will see ubiquitous robotics in my life span.