arXiv subjects

Marco Levorato

Publications and source records attributed to Marco Levorato.

At least 37 records · Page 2Linked to original sources

arXiv2023

SC2 Benchmark: Supervised Compression for Split Computing

With the increasing demand for deep learning models on mobile devices, splitting neural network computation between the device and a more powerful edge server has become an attractive solution. However, existing split computing approaches often underperform compared to a naive baseline of remote computation on compressed data. Recent studies propose learning compressed representations that contain more relevant information for supervised downstream tasks, showing improved tradeoffs between compressed data size and supervised performance. However, existing evaluation metrics only provide an incomplete picture of split computing. This study introduces supervised compression for split computing (SC2) and proposes new evaluation criteria: minimizing computation on the mobile device, minimizing transmitted data size, and maximizing model accuracy. We conduct a comprehensive benchmark study using 10 baseline methods, three computer vision tasks, and over 180 trained models, and discuss various aspects of SC2. We also release sc2bench, a Python package for future research on SC2. Our proposed metrics and package will help researchers better understand the tradeoffs of supervised compression in split computing.

cs.LG↗

arXiv2023

Active Reinforcement Learning for Personalized Stress Monitoring in Everyday Settings

Most existing sensor-based monitoring frameworks presume that a large available labeled dataset is processed to train accurate detection models. However, in settings where personalization is necessary at deployment time to fine-tune the model, a person-specific dataset needs to be collected online by interacting with the users. Optimizing the collection of labels in such phase is instrumental to impose a tolerable burden on the users while maximizing personal improvement. In this paper, we consider a fine-grain stress detection problem based on wearable sensors targeting everyday settings, and propose a novel context-aware active learning strategy capable of jointly maximizing the meaningfulness of the signal samples we request the user to label and the response rate. We develop a multilayered sensor-edge-cloud platform to periodically capture physiological signals and process them in real-time, as well as to collect labels and retrain the detection model. We collect a large dataset and show that the context-aware active learning technique we propose achieves a desirable detection performance using 88\% and 32\% fewer queries from users compared to a randomized strategy and a traditional active learning strategy, respectively.

cs.LG↗

arXiv2022

Matching DNN Compression and Cooperative Training with Resources and Data Availability

To make machine learning (ML) sustainable and apt to run on the diverse devices where relevant data is, it is essential to compress ML models as needed, while still meeting the required learning quality and time performance. However, how much and when an ML model should be compressed, and {\em where} its training should be executed, are hard decisions to make, as they depend on the model itself, the resources of the available nodes, and the data such nodes own. Existing studies focus on each of those aspects individually, however, they do not account for how such decisions can be made jointly and adapted to one another. In this work, we model the network system focusing on the training of DNNs, formalize the above multi-dimensional problem, and, given its NP-hardness, formulate an approximate dynamic programming problem that we solve through the PACT algorithmic framework. Importantly, PACT leverages a time-expanded graph representing the learning process, and a data-driven and theoretical approach for the prediction of the loss evolution to be expected as a consequence of training decisions. We prove that PACT's solutions can get as close to the optimum as desired, at the cost of an increased time complexity, and that, in any case, such complexity is polynomial. Numerical results also show that, even under the most disadvantageous settings, PACT outperforms state-of-the-art alternatives and closely matches the optimal energy cost.

cs.LG↗

arXiv2022

Edge Learning for B5G Networks with Distributed Signal Processing: Semantic Communication, Edge Computing, and Wireless Sensing

To process and transfer large amounts of data in emerging wireless services, it has become increasingly appealing to exploit distributed data communication and learning. Specifically, edge learning (EL) enables local model training on geographically disperse edge nodes and minimizes the need for frequent data exchange. However, the current design of separating EL deployment and communication optimization does not yet reap the promised benefits of distributed signal processing, and sometimes suffers from excessive signalling overhead, long processing delay, and unstable learning convergence. In this paper, we provide an overview on practical distributed EL techniques and their interplay with advanced communication optimization designs. In particular, typical performance metrics for dual-functional learning and communication networks are discussed. Also, recent achievements of enabling techniques for the dual-functional design are surveyed with exemplifications from the mutual perspectives of "communications for learning" and "learning for communications." The application of EL techniques within a variety of future communication systems are also envisioned for beyond 5G (B5G) wireless networks. For the application in goal-oriented semantic communication, we present a first mathematical model of the goal-oriented source entropy as an optimization problem. In addition, from the viewpoint of information theory, we identify fundamental open problems of characterizing rate regions for communication networks supporting distributed learning-and-computing tasks. We also present technical challenges as well as emerging application opportunities in this field, with the aim of inspiring future research and promoting widespread developments of EL in B5G.

cs.IT↗

arXiv2022

BottleFit: Learning Compressed Representations in Deep Neural Networks for Effective and Efficient Split Computing

Although mission-critical applications require the use of deep neural networks (DNNs), their continuous execution at mobile devices results in a significant increase in energy consumption. While edge offloading can decrease energy consumption, erratic patterns in channel quality, network and edge server load can lead to severe disruption of the system's key operations. An alternative approach, called split computing, generates compressed representations within the model (called "bottlenecks"), to reduce bandwidth usage and energy consumption. Prior work has proposed approaches that introduce additional layers, to the detriment of energy consumption and latency. For this reason, we propose a new framework called BottleFit, which, in addition to targeted DNN architecture modifications, includes a novel training strategy to achieve high accuracy even with strong compression rates. We apply BottleFit on cutting-edge DNN models in image classification, and show that BottleFit achieves 77.1% data compression with up to 0.6% accuracy loss on ImageNet dataset, while state of the art such as SPINN loses up to 6% in accuracy. We experimentally measure the power consumption and latency of an image classification application running on an NVIDIA Jetson Nano board (GPU-based) and a Raspberry PI board (GPU-less). We show that BottleFit decreases power consumption and latency respectively by up to 49% and 89% with respect to (w.r.t.) local computing and by 37% and 55% w.r.t. edge offloading. We also compare BottleFit with state-of-the-art autoencoders-based approaches, and show that (i) BottleFit reduces power consumption and execution time respectively by up to 54% and 44% on the Jetson and 40% and 62% on Raspberry PI; (ii) the size of the head model executed on the mobile device is 83 times smaller. We publish the code repository for reproducibility of the results in this study.

cs.LG↗

arXiv2022

Split Computing and Early Exiting for Deep Learning Applications: Survey and Research Challenges

Mobile devices such as smartphones and autonomous vehicles increasingly rely on deep neural networks (DNNs) to execute complex inference tasks such as image classification and speech recognition, among others. However, continuously executing the entire DNN on mobile devices can quickly deplete their battery. Although task offloading to cloud/edge servers may decrease the mobile device's computational burden, erratic patterns in channel quality, network, and edge server load can lead to a significant delay in task execution. Recently, approaches based on split computing (SC) have been proposed, where the DNN is split into a head and a tail model, executed respectively on the mobile device and on the edge server. Ultimately, this may reduce bandwidth usage as well as energy consumption. Another approach, called early exiting (EE), trains models to embed multiple "exits" earlier in the architecture, each providing increasingly higher target accuracy. Therefore, the trade-off between accuracy and delay can be tuned according to the current conditions or application demands. In this paper, we provide a comprehensive survey of the state of the art in SC and EE strategies by presenting a comparison of the most relevant approaches. We conclude the paper by providing a set of compelling research challenges.

eess.SP↗

arXiv2022

SmartDet: Context-Aware Dynamic Control of Edge Task Offloading for Mobile Object Detection

Mobile devices increasingly rely on object detection (OD) through deep neural networks (DNNs) to perform critical tasks. Due to their high complexity, the execution of these DNNs requires excessive time and energy. Low-complexity object tracking (OT) can be used with OD, where the latter is periodically applied to generate "fresh" references for tracking. However, the frames processed with OD incur large delays, which may make the reference outdated and degrade tracking quality. Herein, we propose to use edge computing in this context, and establish parallel OT (at the mobile device) and OD (at the edge server) processes that are resilient to large OD latency. We propose Katch-Up, a novel tracking mechanism that improves the system resilience to excessive OD delay. However, while Katch-Up significantly improves performance, it also increases the computing load of the mobile device. Hence, we design SmartDet, a low-complexity controller based on deep reinforcement learning (DRL) that learns controlling the trade-off between resource utilization and OD performance. SmartDet takes as input context-related information related to the current video content and the current network conditions to optimize frequency and type of OD offloading, as well as Katch-Up utilization. We extensively evaluate SmartDet on a real-world testbed composed of a JetSon Nano as mobile device and a GTX 980 Ti as edge server, connected through a Wi-Fi link. Experimental results show that SmartDet achieves an optimal balance between tracking performance - mean Average Recall (mAR) and resource usage. With respect to a baseline with full Katch-Upusage and maximum channel usage, we still increase mAR by 4% while using 50% less of the channel and 30% power resources associated with Katch-Up. With respect to a fixed strategy using minimal resources, we increase mAR by 20% while using Katch-Up on 1/3 of the frames.

cs.DC↗

arXiv2021

Federated Deep Reinforcement Learning for the Distributed Control of NextG Wireless Networks

Next Generation (NextG) networks are expected to support demanding tactile internet applications such as augmented reality and connected autonomous vehicles. Whereas recent innovations bring the promise of larger link capacity, their sensitivity to the environment and erratic performance defy traditional model-based control rationales. Zero-touch data-driven approaches can improve the ability of the network to adapt to the current operating conditions. Tools such as reinforcement learning (RL) algorithms can build optimal control policy solely based on a history of observations. Specifically, deep RL (DRL), which uses a deep neural network (DNN) as a predictor, has been shown to achieve good performance even in complex environments and with high dimensional inputs. However, the training of DRL models require a large amount of data, which may limit its adaptability to ever-evolving statistics of the underlying environment. Moreover, wireless networks are inherently distributed systems, where centralized DRL approaches would require excessive data exchange, while fully distributed approaches may result in slower convergence rates and performance degradation. In this paper, to address these challenges, we propose a federated learning (FL) approach to DRL, which we refer to federated DRL (F-DRL), where base stations (BS) collaboratively train the embedded DNN by only sharing models' weights rather than training data. We evaluate two distinct versions of F-DRL, value and policy based, and show the superior performance they achieve compared to distributed and centralized DRL.

cs.LG↗

arXiv2021

Spatio-Temporal Split Learning for Autonomous Aerial Surveillance using Urban Air Mobility (UAM) Networks

Autonomous surveillance unmanned aerial vehicles (UAVs) are deployed to observe the streets of the city for any suspicious activities. This paper utilizes surveillance UAVs for the purpose of detecting the presence of a fire in the streets. An extensive database is collected from UAV surveillance drones. With the aid of artificial intelligence (AI), fire stations can swiftly identify the presence of a fire emerging in the neighborhood. Spatio-temporal split learning is applied to this scenario to preserve privacy and globally train a fire classification model. Fires are hazardous natural disasters that can spread very quickly. Swift identification of fire is required to deploy firefighters to the scene. In order to do this, strong communication between the UAV and the central server where the deep learning process occurs is required. Improving communication resilience is integral to enhancing a safe experience on the roads. Therefore, this paper explores the adequate number of clients and data ratios for split learning in this UAV setting, as well as the required network infrastructure.

cs.CV↗

arXiv2021

Supervised Compression for Resource-Constrained Edge Computing Systems

There has been much interest in deploying deep learning algorithms on low-powered devices, including smartphones, drones, and medical sensors. However, full-scale deep neural networks are often too resource-intensive in terms of energy and storage. As a result, the bulk part of the machine learning operation is therefore often carried out on an edge server, where the data is compressed and transmitted. However, compressing data (such as images) leads to transmitting information irrelevant to the supervised task. Another popular approach is to split the deep network between the device and the server while compressing intermediate features. To date, however, such split computing strategies have barely outperformed the aforementioned naive data compression baselines due to their inefficient approaches to feature compression. This paper adopts ideas from knowledge distillation and neural image compression to compress intermediate feature representations more efficiently. Our supervised compression approach uses a teacher model and a student model with a stochastic bottleneck and learnable prior for entropy coding (Entropic Student). We compare our approach to various neural image and feature compression baselines in three vision tasks and found that it achieves better supervised rate-distortion performance while maintaining smaller end-to-end latency. We furthermore show that the learned feature representations can be tuned to serve multiple downstream tasks.

cs.CV↗

arXiv2021

Data Collection and Labeling of Real-Time IoT-Enabled Bio-Signals in Everyday Settings for Mental Health Improvement

Real-time physiological data collection and analysis play a central role in modern well-being applications. Personalized classifiers and detectors have been shown to outperform general classifiers in many contexts. However, building effective personalized classifiers in everyday settings - as opposed to controlled settings - necessitates the online collection of a labeled dataset by interacting with the user. This need leads to several challenges, ranging from building an effective system for the collection of the signals and labels, to developing strategies to interact with the user and building a dataset that represents the many user contexts that occur in daily life. Based on a stress detection use case, this paper (1) builds a system for the real-time collection and analysis of photoplethysmogram, acceleration, gyroscope, and gravity data from a wearable sensor, as well as self-reported stress labels based on Ecological Momentary Assessment (EMA), and (2) collects and analyzes a dataset to extract statistics of users' response to queries and the quality of the collected signals as a function of the context, here defined as the user's activity and the time of the day.

cs.CY↗

arXiv2021

Personalized Stress Monitoring using Wearable Sensors in Everyday Settings

Since stress contributes to a broad range of mental and physical health problems, the objective assessment of stress is essential for behavioral and physiological studies. Although several studies have evaluated stress levels in controlled settings, objective stress assessment in everyday settings is still largely under-explored due to challenges arising from confounding contextual factors and limited adherence for self-reports. In this paper, we explore the objective prediction of stress levels in everyday settings based on heart rate (HR) and heart rate variability (HRV) captured via low-cost and easy-to-wear photoplethysmography (PPG) sensors that are widely available on newer smart wearable devices. We present a layered system architecture for personalized stress monitoring that supports a tunable collection of data samples for labeling, and present a method for selecting informative samples from the stream of real-time data for labeling. We captured the stress levels of fourteen volunteers through self-reported questionnaires over periods of between 1-3 months, and explored binary stress detection based on HR and HRV using Machine Learning Methods. We observe promising preliminary results given that the dataset is collected in the challenging environments of everyday settings. The binary stress detector is fairly accurate and can detect stressful vs non-stressful samples with a macro-F1 score of up to \%76. Our study lays the groundwork for more sophisticated labeling strategies that generate context-aware, personalized models that will empower health professionals to provide personalized interventions.

cs.LG↗

arXiv2021

A Deep Learning Approach to Predict Blood Pressure from PPG Signals

Blood Pressure (BP) is one of the four primary vital signs indicating the status of the body's vital (life-sustaining) functions. BP is difficult to continuously monitor using a sphygmomanometer (i.e. a blood pressure cuff), especially in everyday-setting. However, other health signals which can be easily and continuously acquired, such as photoplethysmography (PPG), show some similarities with the Aortic Pressure waveform. Based on these similarities, in recent years several methods were proposed to predict BP from the PPG signal. Building on these results, we propose an advanced personalized data-driven approach that uses a three-layer deep neural network to estimate BP based on PPG signals. Different from previous work, the proposed model analyzes the PPG signal in time-domain and automatically extracts the most critical features for this specific application, then uses a variation of recurrent neural networks called Long-Short-Term-Memory (LSTM) to map the extracted features to the BP value associated with that time window. Experimental results on two separate standard hospital datasets, yielded absolute errors mean and absolute error standard deviation for systolic and diastolic BP values outperforming prior works.

cs.LG↗

arXiv2021

SeReMAS: Self-Resilient Mobile Autonomous Systems Through Predictive Edge Computing

Edge computing enables Mobile Autonomous Systems (MASs) to execute continuous streams of heavy-duty mission-critical processing tasks, such as real-time obstacle detection and navigation. However, in practical applications, erratic patterns in channel quality, network load, and edge server load can interrupt the task flow execution, which necessarily leads to severe disruption of the system's key operations. Existing work has mostly tackled the problem with reactive approaches, which cannot guarantee task-level reliability. Conversely, in this paper we focus on learning-based predictive edge computing to achieve self-resilient task offloading. By conducting a preliminary experimental evaluation, we show that there is no dominant feature that can predict the edge-MAS system reliability, which calls for an ensemble and selection of weaker features. To tackle the complexity of the problem, we propose SeReMAS, a data-driven optimization framework. We first mathematically formulate a Redundant Task Offloading Problem (RTOP), where a MAS may connect to multiple edge servers for redundancy, and needs to select which server(s) to transmit its computing tasks in order to maximize the probability of task execution while minimizing channel and edge resource utilization. We then create a predictor based on Deep Reinforcement Learning (DRL), which produces the optimum task assignment based on application-, network- and telemetry-based features. We prototype SeReMAS on a testbed composed by a drone, mounting a PixHawk flight controller, a Jetson Nano board, and three 802.11n WiFi interfaces. We extensively evaluate SeReMAS by considering an application where one drone offloads high-resolution images for real-time analysis to three edge servers on the ground. Experimental results show that SeReMAS improves task execution probability by $17\%$ with respect to existing reactive-based approaches.

cs.NI↗

arXiv2020

Neural Compression and Filtering for Edge-assisted Real-time Object Detection in Challenged Networks

The edge computing paradigm places compute-capable devices - edge servers - at the network edge to assist mobile devices in executing data analysis tasks. Intuitively, offloading compute-intense tasks to edge servers can reduce their execution time. However, poor conditions of the wireless channel connecting the mobile devices to the edge servers may degrade the overall capture-to-output delay achieved by edge offloading. Herein, we focus on edge computing supporting remote object detection by means of Deep Neural Networks (DNNs), and develop a framework to reduce the amount of data transmitted over the wireless link. The core idea we propose builds on recent approaches splitting DNNs into sections - namely head and tail models - executed by the mobile device and edge server, respectively. The wireless link, then, is used to transport the output of the last layer of the head model to the edge server, instead of the DNN input. Most prior work focuses on classification tasks and leaves the DNN structure unaltered. Herein, our focus is on DNNs for three different object detection tasks, which present a much more convoluted structure, and modify the architecture of the network to: (i) achieve in-network compression by introducing a bottleneck layer in the early layers on the head model, and (ii) prefilter pictures that do not contain objects of interest using a convolutional neural network. Results show that the proposed technique represents an effective intermediate option between local and edge computing in a parameter region where these extreme point solutions fail to provide satisfactory performance. The code and trained models are available at https://github.com/yoshitomo-matsubara/hnd-ghnd-object-detectors .

cs.CV↗

arXiv2020

Frequency-based Multi Task learning With Attention Mechanism for Fault Detection In Power Systems

The prompt and accurate detection of faults and abnormalities in electric transmission lines is a critical challenge in smart grid systems. Existing methods mostly rely on model-based approaches, which may not capture all the aspects of these complex temporal series. Recently, the availability of data sets collected using advanced metering devices, such as Micro-Phasor Measurement units ($μ$ PMU), which provide measurements at microsecond timescale, boosted the development of data-driven methodologies. In this paper, we introduce a novel deep learning-based approach for fault detection and test it on a real data set, namely, the Kaggle platform for a partial discharge detection task. Our solution adopts a Long-Short Term Memory architecture with attention mechanism to extract time series features, and uses a 1D-Convolutional Neural Network structure to exploit frequency information of the signal for prediction. Additionally, we propose an unsupervised method to cluster signals based on their frequency components, and apply multi task learning on different clusters. The method we propose outperforms the winner solutions in the Kaggle competition and other state of the art methods in many performance metrics, and improves the interpretability of analysis.

cs.LG↗

arXiv2020

Split Computing for Complex Object Detectors: Challenges and Preliminary Results

Following the trends of mobile and edge computing for DNN models, an intermediate option, split computing, has been attracting attentions from the research community. Previous studies empirically showed that while mobile and edge computing often would be the best options in terms of total inference time, there are some scenarios where split computing methods can achieve shorter inference time. All the proposed split computing approaches, however, focus on image classification tasks, and most are assessed with small datasets that are far from the practical scenarios. In this paper, we discuss the challenges in developing split computing methods for powerful R-CNN object detectors trained on a large dataset, COCO 2017. We extensively analyze the object detectors in terms of layer-wise tensor size and model size, and show that naive split computing methods would not reduce inference time. To the best of our knowledge, this is the first study to inject small bottlenecks to such object detectors and unveil the potential of a split computing approach. The source code and trained models' weights used in this study are available at https://github.com/yoshitomo-matsubara/hnd-ghnd-object-detectors .

cs.CV↗

arXiv2020

Data-Driven Path Selection for Real-Time Video Streaming at the Network Edge

In this paper, we present a framework for the dynamic selection of the wireless channels used to deliver information-rich data streams to edge servers. The approach we propose is data-driven, where a predictor, whose output informs the decision making of the channel selector, is built from available data on the transformation imposed by the network on previously transmitted packets. The proposed technique is contextualized to real-time video streaming for immediate processing. The core of our framework is the notion of probes, that is, short bursts of packets transmitted over unused channels to acquire information while generating a controlled impact on other active links. Results indicate a high accuracy of the prediction output and a significant improvement in terms of received video quality when the prediction output is used to dynamically select the used channel for transmission.

cs.NI↗