Dimitris Koutromanos

H2IOSC Training  Building Transformer-Based Natural Language Processing with NVidia  Organized by CNR IAC    July 25, 2024 | h 9:00 – 17:00  ROME, CNR IAC headquarter, Via dei Taurini 19, Ground Floor room    Gain an understanding of how transformers are used as the foundation of modern large language models (LLMs). Learn how to use these models for various Natural Language Processing (NLP) tasks, including text classification, named-entity recognition (NER), author attribution, and question-answering. Additionally, understand how to analyze various model features, constraints, and characteristics to determine which model is best suited for a particular use case based on metrics, domain specificity, and available resources. Find out more about the one-day H2IOSC training here: https://lnkd.in/dP2UMfzF     Consiglio Nazionale delle Ricerche CNR Istituto per le Applicazioni del Calcolo "Mauro Picone"

14

🚀 Exciting times at DeepLearn 2024! 🚀 Professor: Michalis Vazirgiannis University: École Polytechnique Course: Graph Machine Learning and Multimodal Graph Generative AI Level: Intermediate/Advanced Dates: 15th - 19th of July Location: Universidade da Maia - UMAIA 📅 Course Highlights: Graph Machine Learning (GML): Understand the challenges of GML, including graph similarity via graph kernels with applications in community detection and impact evaluation. GraKeL Python Library: Get introduced to the GraKeL library for enabling machine learning with graph kernels. Deep Learning for Graphs: Learn about Graph Neural Networks (GNNs) with node and graph embedding algorithms and their experimental evaluations. Message Passing GNN Model: Focus on the GNN model, while exploring alternatives such as RWNN for explainability and Hyperbolic GNNs for hierarchical relations. Graph Generative AI: Discover the emerging field of Graph Generative AI using deep neural networks and large language models, with applications in biomedicine and molecule generation. Multimodality for Graphs: Explore the topic of multimodality for graphs and its implications in various domains. Join us to dive deep into the fascinating world of graph machine learning and multimodal graph generative AI, and enhance your skills in applying these advanced techniques to real-world data structures. For more details, visit: https://lnkd.in/dxZ-S_EZ #DeepLearn2024 #MachineLearning #GraphML #GraphNeuralNetworks #AI #GraphKernels #TechInnovation #DeepLearning #GenerativeAI #MultimodalGraphs

4

📣 The workshop “Quantum Optics meets Tensor Networks” will be held on September 4-5, 2024 at the Department of Physics in Bari. The workshop targets researchers and scholars interested in the field of Quantum Optics, with a particular focus on Waveguide Quantum Electrodynamics (QED), and experts working in the field of Tensor Network methods. The growing number of geometries and procedures within quantum optical systems and Waveguide QED call for the development of new theoretical models and efficient simulation algorithms. Tensor Networks feature the ideal tool to achieve this goal by their capabilities in building efficient and accurate models and simulations of complex quantum systems. The workshop will gather international experts in theoretical and experimental Quantum Optics and Waveguide QED, and Tensor Network methods, promoting collaborations and cross-fertilization with potential enrichment for both communities. The programme includes scientific talks and open discussions aiming at encourage collaborations and idea exchange. For more information visit: https://lnkd.in/dQ73pteA #QuantumOptics #WaveguideQED #TensorNetworks #Quantum

26

This week's final featured #DigitalDiscovery article is from Andolina and Saidi, and demonstrates the development of deep neural network atomistic potential for melting temperature prediction, trained on generated DFT datasets. Read their full paper here https://lnkd.in/dBRSCRJ3 #openaccess

9

💬 Petros Anastasiadis presented at IEEE Cluster 2024 his work on "Uncut-GEMMs : Communication-aware matrix-multiplication on multi-GPU nodes. 💻 📜 About the Paper: General Matrix Multiplication (GEMM) is a core kernel in HPC and ML often run on multi-GPU clusters. However, existing implementations are outdated and don't fully utilize modern clusters, leading to performance losses. Our work introduces a new static scheduler with advanced algorithmic, communication, and auto-tuning optimizations for multi-GPU GEMM. We've integrated these improvements into an open-source library that achieves an average performance boost of 1.37x and 1.29x over current leading libraries on a system with 8 NVIDIA A100 GPUs, for double and single precision, respectively." 📊 About the picture: The square GEMM (M=N=K) FP64 performance of state-of-the-art libraries in 8 A100 GPUs (system peak = dashed line). 😎 Our work offers robust performance regardless of the data placement, avoids imbalance, and outperforms all previous approaches. #HPC #GPU ICCS - NTUA #EuroHPC EuroHPC Joint Undertaking (EuroHPC JU)

5

In the September #AUGIWORLD, Ghábor Imre Kovàcs succinctly explains different types of AI, their applications, and how they are transforming engineering practices. Check it out on page 34 here: https://smpl.is/9kz6c #AUGI

1

🎉Volume 16, Issue 18 has been released! ✍️Cover paper: Considering the Effects of Horizontal Heterogeneities in #Satellite-Based Large-Scale Statistics of #Cloud Optical Properties by Tamás Várnai and Alexander Marshak 🔗Check out the cover story and full issue: https://brnw.ch/21wNe0x #newissue #issuecover #remotesensing

5

Looking forward to #EuroSys2024 in Athens! Foteini Strati will present our Orion system for fine-grain, interference-aware GPU sharing to improve ML hardware utilization and cost. We find that individual ML jobs often underutilize GPUs as they consist of data-dependent kernels that rarely simultaneously saturate both compute and memory bandwidth on GPUs. To improve GPU utilization while minimizing interference, our key insight is to share GPUs between jobs by scheduling at the fine granularity of individual GPU kernels and taking into account the compute versus memory resource profiles of each kernel. Paper: https://lnkd.in/gbq_wxFr Code: https://lnkd.in/gHkQd-ET

332

3 σχόλια

Interested in differential-algebraic equations (DAEs) and want to contribute to open source? Want to get paid? Check out the #SciML Small Grants program where you can be paid to contribute to contribute to #DAE solvers, nonlinear solvers, and more! For more information, see: https://lnkd.in/dKa_iNqv #julialang #differentialequations #oss #benchmarks #PerformanceBoost

54

1 σχόλιο

New RSLab paper titled "Local-to-Global Cross-Modal Attention-Aware Fusion for HSI-X Semantic Segmentation" on IEEE Transactions on Geoscience and Remote Sensing. Link to the paper: https://lnkd.in/dJhte4wv #RemoteSensing #EarthObservation #FeatureExtraction #DeepLearning #SemanticSegmentation #HyperspectralImage

25

1 σχόλιο

Call for Papers Alert: Special Issue on Revisiting Software Testing Submit your research for an issue of Computer magazine! The issue will showcase the landscape of software for all systems testing in both theory and practice. More information @ https://bit.ly/3LNsiLF #IEEE

9

Call for Papers Alert: Special Issue on Revisiting Software Testing Submit your research for an issue of Computer magazine! The issue will showcase the landscape of software for all systems testing in both theory and practice. More information @ https://bit.ly/3LNsiLF #IEEE

4

This paper investigates the coexistence of #orthogonal #time-#frequency #space (#OTFS) modulation with current #fourth- #and #fifth-#generation (#4G/#5G) communication systems that primarily use #orthogonal #frequency-#division #multiplexing (#OFDM) waveforms. The authors first derive the #input-#output-#relation of OTFS in the considered coexisting system. In this derivation, they consider (i) the inclusion of multiple #cyclic #prefixes (#CPs) with unequal lengths to the OTFS signal and (ii) #edge #carrier #unloading (#ECU), to account for the impacts of CP length, frame structure, and subcarrier arrangement described in #3GPP #standards for 4G/5G systems. Their analysis reveals that the inclusion of multiple CPs to the OTFS signal and ECU lead to the channel response exhibiting spreading effects/leakage along the Doppler and delay dimensions, respectively. Consequently, the effective sampled #delay-#Doppler (#DD) domain channel model for OTFS in coexisting systems may exhibit reduced sparsity.----Akram Shafie, Jinhong Yuan, @Paul Fitzpatrick, Taka Sakurai, @Yuting Fang More details can be found at this link: https://lnkd.in/gYZvCFKe

7

1 σχόλιο