Susinder Rajan Gulasekaran

1. 🚀 TextGrad uses text feedback for automatic differentiation, offering a new method for prompt optimization in large language models, effective in Q&A, molecular optimization, and radiotherapy planning. 2. 🔧 DSPy provides the algorithmic foundation for TextGrad to optimize LLMs' prompts and weights, reducing the need for manual adjustments. 3. 🌐 Automatic differentiation is crucial in deep learning, and TextGrad applies it to the text domain, optimizing through text feedback. 4. ✨ ProTeGi leverages natural language gradients and prompt editing to effectively optimize LLMs' prompts, finding the optimal configuration. 5. ❓ Despite TextGrad's innovation, its practical application may be limited by the performance of the teacher model and the need for extensive API calls. #AI #TextGrad #Innovation #DSPy #LLMOptimization #Algorithm #AutoDiff #DeepLearning #ProTeGi #PromptOptimization #AIOptimization #AIChallenges #TechLimitations

1

💥💥💥 LongRAG: A Dual-Perspective Retrieval-Augmented Generation Paradigm for Long-Context Question Answering Qingfei Zhao, Ruobing Wang, Yukuo Cen, Daren Zha, Shicheng Tan, Yuxiao Dong, Jie Tang Abstract Long-Context Question Answering (LCQA), a challenging task, aims to reason over long-context documents to yield accurate answers to questions. Existing long-context Large Language Models (LLMs) for LCQA often struggle with the "lost in the middle" issue. Retrieval-Augmented Generation (RAG) mitigates this issue by providing external factual evidence. However, its chunking strategy disrupts the global long-context information, and its low-quality retrieval in long contexts hinders LLMs from identifying effective factual details due to substantial noise. To this end, we propose LongRAG, a general, dual-perspective, and robust LLM-based RAG system paradigm for LCQA to enhance RAG's understanding of complex long-context knowledge (i.e., global information and factual details). We design LongRAG as a plug-and-play paradigm, facilitating adaptation to various domains and LLMs. Extensive experiments on three multi-hop datasets demonstrate that LongRAG significantly outperforms long-context LLMs (up by 6.94%), advanced RAG (up by 6.16%), and Vanilla RAG (up by 17.25%). Furthermore, we conduct quantitative ablation studies and multi-dimensional analyses, highlighting the effectiveness of the system's components and fine-tuning strategies. 👉 https://lnkd.in/d7PaajBQ #machinelearning

1

《☆》:: What is the group cohomology of the homotopy and cohomotopy groups of a 𝑑=11 qudit system?《◇¿》. 《●》:: Quantum gates and operations induce group actions on the qudit space. how do these symmetry groups act on homotopy and cohomotopy groups, capturing the algebraic essence of gate operations and their interactions? 《¤》 ||○ Quantum gates generate group actions on the qudit Hilbert space, influencing the topological and algebraic structure; Group cohomology translates the abstract group actions into computable invariants, revealing how operations can be composed and interact.●|| 《■♡》: Topological quantum Algorithms can exploit symmetries captured by group cohomology to optimize operations, reduce resource requirements, or enhance fault tolerance. 《☆♡》 ....《¿¡》:: Group cohomology can identify invariant subspaces or conserved quantities within the qudit space, guiding the design of algorithms that leverage these invariants for computational advantage. :: Cohomology can characterize complex entanglement structures, enabling the development of algorithms that utilize multi-qudit entanglement more effectively. ☆:: Cohomology provides insights into the compatibility and interaction between Clifford and non-Clifford gates, essential for achieving universal quantum computation. : 《■◇:》; Engineering materials to host parafermions with 𝑍_11 symmetry, provides a natural framework for 𝑑= qudit system . 《♤●》 , 《Yegoan ☆》

1

1 Comment

Robotics need to be able to perform scene segmentation with accuracy and resilience against adversarial conditions such as a high number of invalid depth pixels or low-light conditions. In response, we present our work on "Diffusion-based RGB-D Semantic Segmentation with Deformable Attention Transformer". Treating the scene segmentation problem as a generative task, while using the deformable attention transformer as an encoding step to introduce robustness against a high number of invalid depth pixels, the method not only achieves and mostly exceeds state-of-the-art performance in both the NYUv2 and SUN RGBD datasets - it importantly presents a higher level of robustness against images with a large number of invalid pixels in the depth map, low-light scenes or when examining segmentation accuracy on small objects. This work was led by Quang Minh Bui. https://lnkd.in/dsv63WGg Project page: https://lnkd.in/dNXiR7zX Paper: https://lnkd.in/dtue9B9C Code: soon available #robotics #computer #vision #segmentation #diffusion #transformer #RGBD #perception

130

3 Comments

This paper by Anita Kumari Rao, Sandeep Sulakh and Deepa Sinha introduces Ω−(Zn), a family of interconnection networks based on a special set S in Zn, examining their structural properties, including #planarity and connectedness, and exploring relationships between derived #graphs and the spectrum of absorption #Cayleygraph and unitary addition Cayley graph. Read more here: https://lnkd.in/ehfKRSVK

16

📢 Original article Sharing-Vol. 23 No. 7 💕 Title: Exploring Implicit Biological Heterogeneity in ASD Diagnosis Using a Multi-Head Attention Graph Neural Network 🤵 Authors: Hyung-Jun Moon, Sung-Bae Cho,* 🔔 Full Text: https://lnkd.in/gtMHteqU 😎Welcome to your reading! #Parkinsons #ParkinsonsDisease #Neurology #BrainHealth #MovementDisorders #Dopamine #NeurodegenerativeDiseases #Healthcare #Research #Awareness #Treatment #SupportGroups #PatientCare #Therapeutics #MentalHealth #Neuroscience #CureParkinsons #BrainMapping #Neurobiology #BrainInjury #NeuroscienceNews #NeuroImaging

3

And half of the systems profiled where based on STM32

11

NVIDIA's pivot to #AI is powered by its #cuda software, established in 2006 to enable general-purpose applications to run on its #GPU. The ecosystem of software created around their products is efficient, user-friendly, and effective, simplifying complex tasks. As most #AI applications operate on NVIDIA's #cuda software, rewriting for other processors like AMD's M1300, Intel Corporation #Gaudi 3, or Amazon's #trainium is time-consuming and risky for developers. To compete with NVIDIA, not only hardware competitiveness is crucial, but also the development of user-friendly software for optimizing #AI tasks. NVIDIA's #AI chips are top-notch, with their software side being a significant advantage. An alternative to #cuda software is #Triton, supporting AMD, Intel Corporation, and NVIDIA. For instance, Meta has integrated #Triton software into its in-house AI-chip, #MTIA. Another contender is #mojo, emphasizing "No cuda required." While cuda programming can be challenging, #Mojo offers a more accessible option for developers to explore.

3

Transformers Transformers has been a major milestone in the field of NLP and heavily used in Generative AI. There are 3 variants of transformer based models and they are 1. Encoder-only 2. Decoder-only 3. Encoder-Decoder Encoder only models: These are also called as autoencoders and these are pretrained using a technique called masked language modeling. A text with a random masked token is sent to the model to predict the masked token. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't _____ at the sign, you will get a ticket. The model is expected to predict the token (word) "stop" These models use bi-directional representations of the input to better understand the full context of a token. Examples of encoder models: BERT Family models Decoder only models: These models are called autoregressive models and are pretrained using a technique called causal language modeling. These models predict the next token using the previous tokens. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't ______" The model will still try to predict the word "stop", but only based on previous tokens. These models are used for generative tasks, including question answering. Examples of decoder only models: GPT Family, Falcon, LLama models Encoder decoder models: These models are called sequence to sequence models and the pretraining objective varies from model to model. For example the popular FLAN - T5 uses a consecutive multitoken masking called span corruption. For example, consider the text "If you don't stop at the sign, you will get a ticket" The training input that will be passed on the encoder model is "If you don't _____ _______ the sign, you will get a ticket. The model will try to predict the tokens "stop at" These models are good at translation tasks. Examples: T5 family In all the above explanations we just took one sentence of text. The LLMs you see in the market are trained on huge volumes of text available over the internet. #LLM #encoder #decoder #transformer #genai

3

🎥 Webinar Recording: Memory for Autonomous Agents 🧠 The recording of our recent webinar on memary - a fully open-source reference implementation for long-term memory in autonomous agents - is now available. In this session, the authors of memary, Julian Saks, Kevin Li, Seyeong Han provide a deep dive into the project and engaged in a thought-provoking discussion and Q&A session around memory, including a multi-agent future with co-learning. In case you haven’t seen it, memary introduces a novel architecture that combines sequential memory with knowledge graphs. This helps provide personalization in the context window but also maintains a larger KG you can query. If you missed the live webinar or want to revisit it, you can now watch the recording here: https://lnkd.in/gJiJmt4E Check out the source repo too: https://lnkd.in/gbyEzrkB

149

2 Comments

Our ECCE 2023 paper "Influence of symmetric BH loop fluctuations on hysteresis loss by Play model" that JSOL wrote is now available on IEEE Xplore: https://lnkd.in/g6THSzNq This study investigates the robustness of the Play model with errors in input data. The Play model can replicate minor loops and calculate hysteresis loss using FEM analysis. Its accuracy depends on the measurement quality of the symmetric BH loops in the input data. Actual measurements may produce noise and uncertainties due to the measurement quality, so some fluctuations should be accepted. Consequently, it is important to understand the influence of fluctuation on the resulting minor loops. #JMAG #BHloops #PlayModel #hysteresisLoss #measurement JMAG WEB MEMBER (free membership) Registering as a "JMAG Web Member" allows you to use JMAG-Express Online (Motor Design Tool), which enables the ability to design motors online and view selected materials. https://lnkd.in/gTE2q3h5

14

This new article discusses our efforts in delivering tensor parallelism, weight-only quantization, vLLM integration, and the RBLN model zoo. It is definitely worth reading! 👉 https://lnkd.in/gRnG8_pR #TensorParallelism #WeightOnlyQuantization #vLLM #ModelZoo

83