Dynamic hard pruning of Neural Networks at the edge of the internet
Valerio L., Nardini F. M., Passarella A., Perego R.
Neural Networks (NN), although successfully applied to several Artificial Intelligence tasks, are often unnecessarily over-parametrized. In edge/fog computing, this might make their training prohibitive on resource-constrained devices, contrasting with the current trend of decentralizing intelligence from remote data centres to local constrained devices. Therefore, we investigate the problem of training effective NN models on constrained devices having a fixed, potentially small, memory budget. We target techniques that are both resource-efficient and performance effective while enabling significant network compression. Our Dynamic Hard Pruning (DynHP) technique incrementally prunes the network during training, identifying neurons that marginally contribute to the model accuracy. DynHP enables a tunable size reduction of the final neural network and reduces the NN memory occupancy during training. Freed memory is reused by a dynamic batch sizing approach to counterbalance the accuracy degradation caused by the hard pruning strategy, improving its convergence and effectiveness. We assess the performance of DynHP through reproducible experiments on three public datasets, comparing them against reference competitors. Results show that DynHP compresses a NN up to 10 times without significant performance drops (up to 3.5% additional error w.r.t. the competitors), reducing up to 80% the training memory occupancy.Source: Journal of network and computer applications 200 (2022). doi:10.1016/j.jnca.2021.103330
DOI: 10.1016/j.jnca.2021.103330
See at:
ISTI Repository 
| CNR ExploRA 
| www.sciencedirect.com
Distilled neural networks for efficient learning to rank
Nardini F. M., Rulli C., Trani S., Venturini R.
Recent studies in Learning to Rank have shown the possibility to effectively distill a neural network from an ensemble of regression trees. This result leads neural networks to become a natural competitor of tree-based ensembles on the ranking task. Nevertheless, ensembles of regression trees outperform neural models both in terms of efficiency and effectiveness, particularly when scoring on CPU. In this paper, we propose an approach for speeding up neural scoring time by applying a combination of Distillation, Pruning and Fast Matrix multiplication. We employ knowledge distillation to learn shallow neural networks from an ensemble of regression trees. Then, we exploit an efficiency-oriented pruning technique that performs a sparsification of the most computationally-intensive layers of the neural network that is then scored with optimized sparse matrix multiplication. Moreover, by studying both dense and sparse high performance matrix multiplication, we develop a scoring time prediction model which helps in devising neural network architectures that match the desired efficiency requirements. Comprehensive experiments on two public learning-to-rank datasets show that neural networks produced with our novel approach are competitive at any point of the effectiveness-efficiency trade-off when compared with tree-based ensembles, providing up to 4x scoring time speed-up without affecting the ranking quality.Source: IEEE transactions on knowledge and data engineering (Online) (2022): 103330. doi:10.1109/TKDE.2022.3152585
DOI: 10.1109/tkde.2022.3152585
See at:
ieeexplore.ieee.org | ISTI Repository | CNR ExploRA
MAT-Index: an index for fast multiple aspect trajectory similarity measuring
De Souza A. P. R., Renso C., Perego R., Bogorny V.
The semantic enrichment of mobility data with several information sources has led to a new type of movement data, the so-called multiple aspect trajectories. Comparing multiple aspect trajectories is crucial for several analysis tasks such as querying, clustering, similarity, and classification. Multiple aspect trajectory similarity measurement is more complex and computationally expensive, because of the large number and heterogeneous aspects of space, time, and semantics that require a different treatment. Only a few works in the literature focus on optimizing all these dimensions in a single solution, and, to the best of our knowledge, none of them proposes a fast point-to-point comparison. In this article we propose the Multiple Aspect Trajectory Index, an index data structure for optimizing the point-to-point comparison of multiple aspect trajectories, considering its three basic dimensions of space, time, and semantics. Quantitative and qualitative evaluations show a processing time reduction of up to 98.1%.Source: Transactions in GIS (Print) (2022). doi:10.1111/tgis.12889
DOI: 10.1111/tgis.12889
Project(s): MASTER 
See at:
onlinelibrary.wiley.com | ISTI Repository | CNR ExploRA
Ensemble model compression for fast and energy-efficient ranking on FPGAs
Gil-Costa V., Loor F., Molina R., Nardini F. M., Perego R., Trani S.
We investigate novel SoC-FPGA solutions for fast and energy-efficient ranking based on machine-learned ensembles of decision trees. Since the memory footprint of ranking ensembles limits the effective exploitation of programmable logic for large-scale inference tasks, we investigate binning and quantization techniques to reduce the memory occupation of the learned model and we optimize the state-of-the-art ensemble-traversal algorithm for deployment on low-cost, energy-efficient FPGA devices. The results of the experiments conducted using publicly available Learning-to-Rank datasets, show that our model compression techniques do not impact significantly the accuracy. Moreover, the reduced space requirements allow the models and the logic to be replicated on the FPGA device in order to execute several inference tasks in parallel. We discuss in details the experimental settings and the feasibility of the deployment of the proposed solution in a real setting. The results of the experiments conducted show that our FPGA solution achieves performances at the state of the art and consumes from 9 × up to 19.8 × less energy than an equivalent multi-threaded CPU implementation.Source: ECIR 2022 - 44th European Conference on IR Research, pp. 260–273, Stavanger, Norway, 10-14/04/2022
DOI: 10.1007/978-3-030-99736-6_18
See at:
doi.org | link.springer.com | CNR ExploRA
A dependency-aware utterances permutation strategy to improve conversational evaluation
Faggioli G., Ferrante M., Ferro N., Perego R., Tonellotto N.
The rapid growth in the number and complexity of conversational agents has highlighted the need for suitable evaluation tools to describe their performance. The main evaluation paradigms move from analyzing conversations where the user explores information needs following a scripted dialogue with the agent. We argue that this is not a realistic setting: different users ask different questions (and in a diverse order), obtaining distinct answers and changing the conversation path. We analyze what happens to conversational systems performance when we change the order of the utterances in a scripted conversation while respecting temporal dependencies between them. Our results highlight that the performance of the system widely varies. Our experiments show that diverse orders of utterances determine completely different rankings of systems by performance. The current way of evaluating conversational systems is thus biased. Motivated by these observations, we propose a new evaluation approach based on dependency-aware utterance permutations to increase the power of our evaluation tools.Source: ECIR 2022 - 44th European Conference on IR Research, pp. 184–198, Stavanger, Norway, 10-14/04/2022
DOI: 10.1007/978-3-030-99736-6_13
See at:
ISTI Repository | doi.org | link.springer.com | CNR ExploRA
The Istella22 dataset: bridging traditional and neural learning to rank evaluation
Dato D., Macavaney S., Nardini F. M., Perego R., Tonellotto N.
Neural approaches that use pre-trained language models are effective at various ranking tasks, such as question answering and ad-hoc document ranking. However, their effectiveness compared to feature-based Learning-to-Rank (LtR) methods has not yet been well-established. A major reason for this is because present LtR benchmarks that contain query-document feature vectors do not contain the raw query and document text needed for neural models. On the other hand, the benchmarks often used for evaluating neural models, e.g., MS MARCO, TREC Robust, etc., provide text but do not provide query-document feature vectors. In this paper, we present Istella22, a new dataset that enables such comparisons by providing both query/document text and strong query-document feature vectors used by an industrial search engine. The dataset consists of a comprehensive corpus of 8.4M web documents, a collection of query-document pairs including 220 hand-crafted features, relevance judgments on a 5-graded scale, and a set of 2,198 textual queries used for testing purposes. Istella22 enables a fair evaluation of traditional learning-to-rank and transfer ranking techniques on the same data. LtR models exploit the feature-based representations of training samples while pre-trained transformer-based neural rankers can be evaluated on the corresponding textual content of queries and documents. Through preliminary experiments on Istella22, we find that neural re-ranking approaches lag behind LtR models in terms of effectiveness. However, LtR models identify the scores from neural models as strong signals.Source: SIGIR '22 - 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 3099–3107, Madrid, Spain, 11-15/07/2022
DOI: 10.1145/3477495.3531740
See at:
ISTI Repository | dl.acm.org | CNR ExploRA
ILMART: interpretable ranking with constrained LambdaMART
Lucchese C., Nardini F. M., Orlando S., Perego R., Veneri A.
Interpretable Learning to Rank (LtR) is an emerging field within the research area of explainable AI, aiming at developing intelligible and accurate predictive models. While most of the previous research efforts focus on creating post-hoc explanations, in this paper we investigate how to train effective and intrinsically-interpretable ranking models. Developing these models is particularly challenging and it also requires finding a trade-off between ranking quality and model complexity. State-of-the-art rankers, made of either large ensembles of trees or several neural layers, exploit in fact an unlimited number of feature interactions making them black boxes. Previous approaches on intrinsically-interpretable ranking models address this issue by avoiding interactions between features thus paying a significant performance drop with respect to full-complexity models. Conversely, ILMART, our novel and interpretable LtR solution based on LambdaMART, is able to train effective and intelligible models by exploiting a limited and controlled number of pairwise feature interactions. Exhaustive and reproducible experiments conducted on three publicly-available LtR datasets show that ILMART outperforms the current state-of-the-art solution for interpretable ranking of a large margin with a gain of nDCG of up to 8%.Source: SIGIR '22 - 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 2255–2259, Madrid, Spain, 11-15/07/2022
DOI: 10.1145/3477495.3531840
See at:
ISTI Repository | dl.acm.org | CNR ExploRA
Fast and compact set intersection through recursive universe partitioning
Pibiri G. E.
We present a data structure that encodes a sorted integer sequence in small space allowing, at the same time, fast intersection operations. The data layout is carefully designed to exploit word-level parallelism and SIMD instructions, hence providing good practical performance. The core algorithmic idea is that of recursive partitioning the universe of representation: a markedly different paradigm than the widespread strategy of partitioning the sequence based on its length. Extensive experimentation and comparison against several competitive techniques shows that the proposed solution embodies an improved space/time trade-off for the set intersection problem.Source: IEEE Data Compression Conference, Online Conference, 23-26/03/2021
See at:
ISTI Repository | CNR ExploRA
Rank/select queries over mutable bitmaps
Pibiri G. E., Kanda S.
The problem of answering rank/select queries over a bitmap is of utmost importance for many succinct data structures. When the bitmap does not change, many solutions exist in the theoretical and practical side. In this work we consider the case where one is allowed to modify the bitmap via a flip(i) operation that toggles its i-th bit. By adapting and properly extending some results concerning prefix-sum data structures, we present a practical solution to the problem, tailored for modern CPU instruction sets. Compared to the state-of-the-art, our solution improves runtime with no space degradation. Moreover, it does not incur in a significant runtime penalty when compared to the fastest immutable indexes, while providing even lower space overhead.Source: Information systems (Oxf.) (2021).
Project(s): BigDataGrapes
See at:
ISTI Repository | CNR ExploRA
Compressed indexes for fast search of semantic data
Perego R., Pibiri G. E., Venturini R.
The sheer increase in volume of RDF data demands efficient solutions for the triple indexing problem, that is devising a compressed data structure to compactly represent RDF triples by guaranteeing, at the same time, fast pattern matching operations. This problem lies at the heart of delivering good practical performance for the resolution of complex SPARQL queries on large RDF datasets. We propose a trie-based index layout to solve the problem and introduce two novel techniques to reduce its space of representation for improved effectiveness. The extensive experimental analysis reveals that our best space/time trade-off configuration substantially outperforms existing solutions at the state-of-the-art, by taking 30-60% less space and speeding up query execution by a factor of 2-81 times.Source: ICDE 2021 - 37th IEEE International Conference on Data Engineering, pp. 2325–2326, Online conference, 19-22/04/2021
DOI: 10.1109/icde51399.2021.00248
Project(s): BigDataGrapes
See at:
ISTI Repository | ieeexplore.ieee.org | CNR ExploRA
PTHash: revisiting FCH minimal perfect hashing
Pibiri G. M., Trani R.
Given a set S of n distinct keys, a function f that bijectively maps the keys of S into the range {0,...,n-1} is called a minimal perfect hash function for S. Algorithms that find such functions when n is large and retain constant evaluation time are of practical interest; for instance, search engines and databases typically use minimal perfect hash functions to quickly assign identifiers to static sets of variable-length keys such as strings. The challenge is to design an algorithm which is efficient in three different aspects: time to find f (construction time), time to evaluate f on a key of S (lookup time), and space of representation for f . Several algorithms have been proposed to trade-off between these aspects. In 1992, Fox, Chen, and Heath (FCH) presented an algorithm at SIGIR providing very fast lookup evaluation. However, the approach received little attention because of its large construction time and higher space consumption compared to other subsequent techniques. Almost thirty years later we revisit their framework and present an improved algorithm that scales well to large sets and reduces space consumption altogether, without compromising the lookup time. We conduct an extensive experimental assessment and show that the algorithm finds functions that are competitive in space with state-of-the art techniques and provide 2 - 4X better lookup time.Source: ACM Conference on Research and Development in Information Retrieval, Canada, Montreal (Virtual Event), 11-15/07/2021
DOI: 10.1145/3404835.3462849
See at:
ISTI Repository | CNR ExploRA
MICROS: Mixed-Initiative ConveRsatiOnal Systems Workshop
Mele I., Muntean C. I., Aliannejadi M., Voskarides N.
The 1st edition of the workshop on Mixed-Initiative ConveRsatiOnal Systems (MICROS@ECIR2021) aims at investigating and collecting novel ideas and contributions in the field of conversational systems. Oftentimes, the users fulfill their information need using smartphones and home assistants. This has revolutionized the way users access online information, thus posing new challenges compared to traditional search and recommendation. The first edition of MICROS will have a particular focus on mixed-initiative conversational systems. Indeed, conversational systems need to be proactive, proposing not only answers but also possible interpretations for ambiguous or vague requests.Source: ECIR 2021 - 43rd European Conference on IR Research, pp. 710–713, Online Conference, March 28 - April 1, 2021
DOI: 10.1007/978-3-030-72240-1_86
See at:
arXiv.org e-Print Archive | ISTI Repository | link.springer.com | link.springer.com | CNR ExploRA
Data Science Workflows for the Cloud/Edge Computing Continuum
Grossi V., Trasarti R., Dazzi P.
Research infrastructures play a crucial role in the development of data science. In fact, the conjunction of data, infrastructures and analytical methods enable multidisciplinary scientists and innovators to extract knowledge and to make the knowledge and experiments reusable by the scientific community, innovators providing an im- pact on science and society. Resources such as data and methods, help domain and data scientists to transform research in an innovation question into a responsible data-driven analytical process. On the other hand, Edge computing is a new computing paradigm that is spreading and developing at an incredible pace. Edge computing is based on the assumption that for certain applications is beneficial to bring the computation as closer as possible to data or end-users. This paper introduces an approach for writing data science workflows targeting research infrastructures that encompass resources located at the edge of the network.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, Virtual Event, Sweden, 25/06/2021
DOI: 10.1145/3452369.3463820
Project(s): SoBigData-PlusPlus
See at:
ISTI Repository | CNR ExploRA
Cloud and Data Federation in MobiDataLab
Carlini E., Dazzi P., Lettich F., Perego R., Renso C.
Today's innovative digital services dealing with the mobility of per- sons and goods produce huge amount of data. To propose advanced and efficient mobility services, the collection and aggregation of new sources of data from various producers are necessary. The overall objective of the MobiDataLab H2020 project is to propose to the mobility stakeholders (transport organising authorities, operators, industry, government and innovators) reproducible methodologies and sustainable tools that foster the development of a data-sharing culture in Europe and beyond. This short paper introduces the key concepts driving the design and definition of the Cloud and Data Federation that stands at the basis of MobiDataLab.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, Virtual Event, Sweden, 25/06/2021
DOI: 10.1145/3452369.3463819
Project(s): ACCORDION
See at:
ISTI Repository | CNR ExploRA
Inter-operability and Orchestration in Heterogeneous Cloud/Edge Resources: The ACCORDION Vision
Korontanis I., Tserpes K., Pateraki M., Blasi L., Violos J., Ferran D., Marin E., Kourtellis N., Coppola M., Carlini E., Ledwo? Z., Tarkowski P., Loven T., González Rozas Y., Kentros M., Dodis M., Dazzi P.
This paper introduces the ACCORDION framework, a novel frame- work for the management of the cloud-edge continuum, targeting the support of NextGen applications with strong QoE requirements. The framework addresses the need for an ever expanding and het- erogeneous pool of edge resources in order to deliver the promise of ubiquitous computing to the NextGen application clients. This endeavor entails two main technical challenges. First, to assure interoperability when incorporating heterogeneous infrastructures in the pool. Second, the management of the largely dynamic pool of edge nodes. The optimization of the delivered QoE stands as the core driver to this work, therefore its monitoring and modelling comprises a core part of the conducted work. The paper discusses the main pillars that support the ACCORDION vision, and provide a description of the three planned use case that are planned to demonstrate ACCORDION capabilities.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, Virtual Event, Sweden, 25/06/2021
DOI: 10.1145/3452369.3463816
Project(s): ACCORDION , ACCORDION
See at:
ISTI Repository | CNR ExploRA | ZENODO
Collaborative Visual Environments for Evidence Taking in Digital Justice: a Design Concept
Erra U., Capece N., Lettieri N., Fabiani E., Banterle F., Cignoni P., Dazzi P., Aleotti J., Monica R.
In recent years, Spatial Computing (SC) has emerged as a novel paradigm thanks to the advancements in Extended Reality (XR), remote sensing, and artificial intelligence. Computers are nowadays more and more aware of physical environments (i.e. objects shape, size, location and movement) and can use this knowledge to blend technology into reality seamlessly, merge digital and real worlds, and connect users by providing innovative interaction methods. Criminal and civil trials offer an ideal scenario to exploit Spatial Computing. The taking of evidence, indeed, is a complex activity that not only involves several actors (judges, lawyers, clerks, advi- sors) but it often requires accurate topographic surveys of places and objects. Moreover, another essential means of proof, the "judi- cial experiments" - reproductions of real-world events (e.g. a road accident) the judge uses to evaluate if and how a given fact has taken place - could be usefully carried out in virtual environments. In this paper we propose a novel approach for digital justice based on a multi-user, multimodal virtual collaboration platform that enables technology-enhanced acquisition and analysis of trial evidence.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, Sweden, Virtual Event, 25/06/2021
DOI: 10.1145/3452369.3463820
Project(s): ACCORDION
See at:
ISTI Repository | CNR ExploRA
ACCORDION: Edge Computing for NextGen Applications
Dazzi P.
Cloud computing has played a vital role in the digital revolution. Clouds enable consumers and businesses to use applications without dealing with local installations and the associated complexity. However, a big class of applications is currently being blocked because of their dependency on on-site infrastructures or specialised end-devices but also because they are too latency-sensitive or data-dependent to be moved to the public cloud.Source: ERCIM news online edition 125 (2021).
Project(s): ACCORDION
See at:
ercim-news.ercim.eu | ISTI Repository | CNR ExploRA
Latency Preserving Self-optimizing Placement at the Edge
Ferrucci L., Mordacchini M., Coppola M., Carlini E., Kavalionak H., Dazzi P.
The Internet is experiencing a fast expansion at its edges. The wide availability of heterogeneous resources at the Edge is pivotal in the definition and extension of traditional Cloud solutions toward supporting the development of new applications. However, the dynamic and distributed nature of these resources poses new challenges for the optimization of the behaviour of the system. New decentralized and self-organizing methods are needed to face the needs of the Edge/Cloud scenario and to optimize the exploitation of Edge resources. In this paper we propose a distributed and adaptive solution that reduces the number of replicas of application services that are executed throughout the system, all the while ensuring that the latency constraints of applications are met, thus allowing to also meet the end users' QoS requirements. Experimental evaluations through simulation show the effectiveness of the proposed approach.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, Virtual Event, Sweden, 25/06/2021
DOI: 10.1145/3452369.3463815
Project(s): ACCORDION , ACCORDION
See at:
ISTI Repository | CNR ExploRA | ZENODO
An Osmotic Ecosystem for Data Streaming Applications in Smart Cities
Carlini E., Carnevale L., Coppola M., Dazzi P., Mencagli G., Talia D., Villari M.
Modern multi-tier Cloud-Edge-IoT computational platforms seamlessly map with the distributed and hierarchical nature of smart cities infrastructure. However, classical tools and methodologies to organise data as well as computational and network resources are poorly equipped to tackle the dynamic and heterogeneous environments of smart cities. In this paper we propose a reference architecture that aims to establish a unified approach for the orchestration of modern Cloud-Edge-IoT infrastructures and resources specifically tailored for data streaming applications in smart-cities. Stemming from the proposed reference architecture, we also discuss a series of open challenges, which we believe represent relevant research directions in the nearest future.Source: FRAME'21 - 1st Workshop on Flexible Resource and Application Management on the Edge, online, Sweden, 25/06/2021
DOI: 10.1145/3452369.3463822
See at:
CNR ExploRA
ACCORDION: Architecture finally defined
Dazzi P.
The H2020 project ACCORDION will very soon deliver the core modules of its novel platform, with a complete release of the integrated platform to follow soon after.Project(s): ACCORDION
See at: CNR ExploRA | www.hipeac.net