New Ways of Software Development
Xinyi Zhang (Huawei)

Abstract There are a lot of technologies around Integrated Development Environment (aka IDE) for decades, and they keep evolving to support new platforms, achieve new business goals, etc. In this talk, I’d like to introduce what problems IDE Lab is trying to solve today, the approaches we are taking, and challenges we are facing with.

Bio Xinyi Zhang is Director of IDE Lab, Huawei. Before joining Huawei, Xinyi had been working in Microsoft for 13 years, mainly focusing on Visual Studio, Visual Studio Code, and developer ecosystems.

Scoring Neural Networks Without Training Them 
Amos Storkey (UoE)

Abstract Neural Architecture search is a costly and often cumbersome business. There are many architectures that perform poorly, and often more noise than signal for architectures that perform well. In this talk I will discuss a couple of approaches we have recently developed that look at key information from neural networks pre-learning that given strong indicators of final network performance, and describe the use in hardware-aware network learning and neural architecture search.

Work is in conjunction with Joseph Mellor, Jack Turner, Elliot J. Crowley and Gavin Gray.

Bio  Amos Storkey is Professor of Machine Learning and AI in the School of Informatics, University of Edinburgh, and has over 25 years history in neural network research. He now leads a research team focused on deep neural networks, transfer learning, Bayesian methods, transactional and distributed machine learning, and efficient learning and inference. He is currently also director of the EPSRC Centre for Doctoral Training in Data Science,

The past, present and future of openGauss
Jinwei Zhu (Huawei)

Abstract openGauss is an open source relational database management system that is released with the Mulan PSL v2. with the kernel derived from PostgreSQL, openGauss is built on Huawei’s years of experience in the database field and continuously provides competitive features tailored to enterprise-grade scenarios.

Bio  Jinwei Zhu is Project Manager of openGauss Kernel Team.  He has a PhD in Mathematics, he graduated from the Chinese Academy of Sciences and has focused on databases for about 7 years.

Neural and neuronal networks, or how does the brain compute and learn?
Mattias Hennig (UoE)

Abstract Artificial neural networks were initially conceived as models to understand learning and computation in the brain. At the same time, they have been and are still frequently dismissed by neuroscientists as overly simplistic and unrealistic. In reality perhaps both sides have valid points: artificial networks can yield valuable analytic understanding, but their limitations have to be exposed clearly and rectified to make progress in understanding brain function. Drawing on recent research in neuroscience, I will explore some parts of this Venn diagram and give a perspective on how the two fields can interact to mutual benefit.

Bio Matthias Hennig is a Reader in Computational Neuroscience at the School of Informatics, University of Edinburgh. He studied Physics at Bochum University and obtained a PhD in Computational Neuroscience at the University of Stirling in 2005. Following postdoctoral training through a MRC Training Fellowship and a MRC Career Development Award, he was appointed lecturer at Informatics in 2012. His research interest is focussed on development, stability and processing in neuronal circuits, using computational and mathematical modelling. To enable links between theory and experimental biology, his group also develops methods and open-source tools for analysis and interpretation of large scale neural recordings, and develops and advocates open science approaches. He currently serves as the director of the Institute for Adaptive and Neural Computation in Informatics.

Technical Challenges in Huawei Petal Search
Zhonghua Li  (Huawei)

Abstract In 2020, Huawei launched its official search engine app — Petal Search, and it is now available in over 170 countries and regions and supports over 50 languages, letting users conveniently and instantly find out the information and services they need. Petal Search offers over 20 kinds of search capabilities (including web, app, news, videos, images, shopping, flights, etc.) and various tools (such as weather, calculators) to improve users search experience. However, as a young and fast growing search engine, Petal Search is also facing many technical challenges .This talk will share some of these technical challenges and progress in a few target domains.

Bio Zhonghua Li received her Ph.D. degree in Computer Science from National University of Singapore (NUS) in July 2013. She got her bachelor in Computer Science from Northwestern Polytechnical University in 2008. After her Ph.D., she worked as a data scientist in semantic retrieval and recommendation in industry for four years. Then she joined the Poisson Lab, 2012, Huawei in 2017 and now is a Technical Expert leading the ranking team in Poisson Lab. She is a one of the main contributors of the ranking system of Huawei Petal Search. Her research interests include multimedia information retrieval, recommendation, and data mining.

Zero-Shot Cross-lingual Semantic Parsing   
Mirella Lapata (UoE)

Abstract Semantic parsing is the task of mapping natural language utterances to machine-interpretable expressions such as SQL or a logical meaning representation. It has emerged as a key technology for developing natural language interfaces, especially in the context of question answering where a semantically complex question is mapped to an executable query to retrieve an answer, or denotation. Datasets for semantic parsing scarcely consider languages other than English and professional translation can be prohibitively expensive. Recent work has successfully applied machine translation to localize parsers to new languages. However, high-quality machine translation is less viable for lower resource languages, and can introduce performance limiting artifacts, struggling to accurately model native speakers. We study cross-lingual semantic parsing as a zero-shot problem without parallel data for new languages. We propose a multi-task encoder-decoder model to transfer parsing knowledge to additional languages using only English-Logical form paired data and unlabeled, mono-lingual utterances in each target language. Our encoder learns language-agnostic representations and is jointly optimized for generating logical forms or utterance reconstruction and against language discriminability. Our zero-shot parser performs above back-translation baselines and, in some cases, approaches the supervised upper bound.

Bio  Mirella Lapata is professor of natural language processing in the School of Informatics at the University of Edinburgh. Her research focuses on getting computers to understand, reason with, and generate natural language. She is the first recipient (2009) of the British Computer Society and Information Retrieval Specialist Group (BCS/IRSG) Karen Sparck Jones award, a Fellow of the ACL and the Royal Society of Edinburgh. She has also received best paper awards in leading NLP conferences and has served on the editorial boards of the Journal of Artificial Intelligence Research, the Transactions of the ACL, and Computational Linguistics. She was president of SIGDAT (the group that organizes EMNLP) in 2018.

Operating Systems architecture for emerging non-traditional hardware topologies     
Antonio Barbalace (UoE)

Abstract Today’s computer hardware is increasingly heterogeneous, including several special purpose and reconfigurable accelerators along with the central processing unit (CPU). Emerging platforms go a step further including processing units (CPUs and/or accelerators), in the storage, network, and memory hierarchies (near data processing architectures). Therefore, introducing hardware topologies that were not used before — non-traditional, e.g., a single computer with multiple diverse CPUs, GPUs, etc. Operating systems were born out of the necessity of providing a generic interface for software to run on a single platform while exploiting all available hardware resources. With emerging hardware, traditional software architectures enforce running one operating system per processing unit, or at least per CPU complex. Hence, a single computer runs multiple Operating systems (OSes) — which is against the initial purpose OSes were built for. This talk argues for a new operating systems architecture for emerging non-traditional hardware — an architecture that gives to the Operating system the role that it was developed for.

Bio Antonio Barbalace is a Senior Lecturer (Associate Professor) at the School of Informatics of the University of Edinburgh, Scotland. Before, he was an Assistant Professor in the Computer Science Department, at Stevens Institute of Technology, New Jersey. Prior to that, he was a Principal Research Scientist and Manager at Huawei’s Munich Research Center (MRC). He was a Research Assistant Professor, and before a Postdoc, within the Systems Software Research Group, ECE Department, at Virginia Tech, Virginia. He earned a PhD in Industrial Engineering from the University of Padova, Italy, and an MS and BS in Computer Engineering from the same University. Antonio Barbalace’s research interests include all aspects of system software, embracing hypervisors, operating systems, runtime systems, and compilers/linkers, for emerging highly-parallel and heterogeneous computer architectures, including all-types of Near-Data Processing (NDP) platforms.

Modelling indoor spatial data for localisation, navigation and search
Firas Alsehly (Huawei)

Abstract On the verge of recent growth in maps, navigation and data driven applications, geospatial data available indoors became exceptionally valuable. As various technologies emerges from outdoor to indoor spaces, data representation sees a gap in the lack of equivalent context such as road networks. However, as humans we tend to move with more agility when indoor and the variation in spatially connected location data doesn’t follow as crisp representation as of outdoors. However, topological nested graphs can emulate data boundaries and reconstruct indoor skeleton.

This talk discuss methods for constructing a topology and connectivity of the indoor structure represented by a connected, embedded graph to hold geometric and semantic information of multi-floor crowded buildings. It also demonstrate how such modelling can drive significant improvements for indoor localisation and navigation applications via trajectory correction, global optimisation and autonomous calibration.

Bio Firas is the leader of Indoor Positioning and Navigation Lab in Huawei Edinburgh Research Center. He completed his PhD in the University of Edinburgh in signal processing and adaptive systems and sums over 10 years of R&D experience in geospatial data analysis delivering global scale positioning algorithms and data pipeline for crowd-sourcing.

Operating System in Cloud Native era
Xinwei Hu  (Huawei)

Abstract There are 2 stages for enterprise applications to migrate to the Cloud. First stage is “on cloud”. In this stage, cloud vendors focus on providing solution to make it easier for legacy application, so that they can run on the cloud without modification. Second stage will be “in cloud”. In this stage, enterprise applications need to be changed to fully adapt what cloud provides, and be truly “cloud native”. We think there are opportunities for Operating System to help applications achieve “in cloud”.

Bio Xinwei Hu is Chief ICT OS Architect.  He is an experienced Chief System Architect with a demonstrated history of working in the telecommunications industry. Strong engineering professional skilled in Open Source Software, Python, C++, Bash, and Unix.