A5031344912- Advanced Image and Video Retrieval Techniques
- Algorithms and Data Compression
- Image Retrieval and Classification Techniques
- Web Data Mining and Analysis
- Evaluation and Optimization Models
- Advanced Data Storage Technologies
- Topic Modeling
- Multimodal Machine Learning Applications
- Underwater Vehicles and Communication Systems
- Evaluation Methods in Various Fields
- Plasma Diagnostics and Applications
- Advanced Decision-Making Techniques
- Underwater Acoustics Research
- Parallel Computing and Optimization Techniques
ETH Zurich
2023-2025
Trends in hardware, the prevalence of cloud, and rise highly demanding applications have ushered an era specialization that quickly changes how data is processed at scale. These are likely to continue accelerate next years as new technologies adopted deployed: smart NICs, storage, memory, disaggregated specialized accelerators (GPUS, TPUs, FPGAs), a wealth ASICs specifically created deal with computationally expensive tasks (e.g., cryptography or compression). In this tutorial, we focus on...
Recent breakthroughs in large models have highlighted the critical significance of data scale, labels and modals. In this paper, we introduce MS MARCO Web Search, first large-scale information-rich web dataset, featuring millions real clicked query-document labels. This dataset closely mimics real-world document query distribution, provides rich information for various kinds downstream tasks encourages research areas, such as generic end-to-end neural indexer models, embedding next...
Vector search has emerged as the foundation for large-scale information retrieval and machine learning systems, with engines like Google Bing processing tens of thousands queries per second on petabyte-scale document datasets by evaluating vector similarities between encoded query texts web documents. As performance demands systems surge, accelerated hardware offers a promising solution in post-Moore's Law era. We introduce FANNS, an end-to-end scalable framework FPGAs. Given user-provided...
Retrieval-augmented generation (RAG) can enhance the quality of large language models (LLMs) by incorporating external token databases. However, retrievals from databases constitute a substantial portion overall time, particularly when are periodically performed to align retrieved content with latest states generation. In this paper, we introduce PipeRAG, novel algorithm-system co-design approach reduce latency and quality. PipeRAG integrates (1) pipeline parallelism enable concurrent...
A Retrieval-Augmented Language Model (RALM) combines a large language model (LLM) with vector database to retrieve context-specific knowledge during text generation. This strategy facilitates impressive generation quality even smaller models, thus reducing computational demands by orders of magnitude. To serve RALMs efficiently and flexibly, we propose Chameleon , heterogeneous accelerator system integrating both LLM search accelerators in disaggregated architecture. The heterogeneity...
Vector search has emerged as the foundation for large-scale information retrieval and machine learning systems, with engines like Google Bing processing tens of thousands queries per second on petabyte-scale document datasets by evaluating vector similarities between encoded query texts web documents. As performance demands systems surge, accelerated hardware offers a promising solution in post-Moore's Law era. We introduce \textit{FANNS}, an end-to-end scalable framework FPGAs. Given...