Lothar Thiele

Fabian Mager, Dominik Baumann, Romain Jacob et al.

Closing feedback loops fast and over long distances is key to emerging applications; for example, robot motion control and swarm coordination require update intervals of tens of milliseconds. Low-power wireless technology is preferred for its low cost, small form factor, and flexibility, especially if the devices support multi-hop communication. So far, however, feedback control over wireless multi-hop networks has only been shown for update intervals on the order of seconds. This paper presents a wireless embedded system that tames imperfections impairing control performance (e.g., jitter and message loss), and a control design that exploits the essential properties of this system to provably guarantee closed-loop stability for physical processes with linear time-invariant dynamics. Using experiments on a cyber-physical testbed with 20 wireless nodes and multiple cart-pole systems, we are the first to demonstrate and evaluate feedback control and coordination over wireless multi-hop networks for update intervals of 20 to 50 milliseconds.

Wenying Duan, Xiaoxi He, Zimu Zhou et al.

Spatial-temporal graph models are prevailing for abstracting and modelling spatial and temporal dependencies. In this work, we ask the following question: whether and to what extent can we localise spatial-temporal graph models? We limit our scope to adaptive spatial-temporal graph neural networks (ASTGNNs), the state-of-the-art model architecture. Our approach to localisation involves sparsifying the spatial graph adjacency matrices. To this end, we propose Adaptive Graph Sparsification (AGS), a graph sparsification algorithm which successfully enables the localisation of ASTGNNs to an extreme extent (fully localisation). We apply AGS to two distinct ASTGNN architectures and nine spatial-temporal datasets. Intriguingly, we observe that spatial graphs in ASTGNNs can be sparsified by over 99.5\% without any decline in test accuracy. Furthermore, even when ASTGNNs are fully localised, becoming graph-less and purely temporal, we record no drop in accuracy for the majority of tested datasets, with only minor accuracy deterioration observed in the remaining datasets. However, when the partially or fully localised ASTGNNs are reinitialised and retrained on the same data, there is a considerable and consistent drop in accuracy. Based on these observations, we reckon that \textit{(i)} in the tested data, the information provided by the spatial dependencies is primarily included in the information provided by the temporal dependencies and, thus, can be essentially ignored for inference; and \textit{(ii)} although the spatial dependencies provide redundant information, it is vital for the effective training of ASTGNNs and thus cannot be ignored during training. Furthermore, the localisation of ASTGNNs holds the potential to reduce the heavy computation overhead required on large-scale spatial-temporal data and further enable the distributed deployment of ASTGNNs.

Zhongnan Qu, Zimu Zhou, Yongxin Tong et al.

Data collected by IoT devices are often private and have a large diversity across users. Therefore, learning requires pre-training a model with available representative data samples, deploying the pre-trained model on IoT devices, and adapting the deployed model on the device with local data. Such an on-device adaption for deep learning empowered applications demands data and memory efficiency. However, existing gradient-based meta learning schemes fail to support memory-efficient adaptation. To this end, we propose p-Meta, a new meta learning method that enforces structure-wise partial parameter updates while ensuring fast generalization to unseen tasks. Evaluations on few-shot image classification and reinforcement learning tasks show that p-Meta not only improves the accuracy but also substantially reduces the peak dynamic memory by a factor of 2.5 on average compared to state-of-the-art few-shot adaptation methods.

Zexin Li, Xiaoxi He, Yufei Li et al.

Future intelligent robots are expected to process multiple inputs simultaneously (such as image and audio data) and generate multiple outputs accordingly (such as gender and emotion), similar to humans. Recent research has shown that multi-input single-output (MISO) deep neural networks (DNN) outperform traditional single-input single-output (SISO) models, representing a significant step towards this goal. In this paper, we propose MIMONet, a novel on-device multi-input multi-output (MIMO) DNN framework that achieves high accuracy and on-device efficiency in terms of critical performance metrics such as latency, energy, and memory usage. Leveraging existing SISO model compression techniques, MIMONet develops a new deep-compression method that is specifically tailored to MIMO models. This new method explores unique yet non-trivial properties of the MIMO model, resulting in boosted accuracy and on-device efficiency. Extensive experiments on three embedded platforms commonly used in robotic systems, as well as a case study using the TurtleBot3 robot, demonstrate that MIMONet achieves higher accuracy and superior on-device efficiency compared to state-of-the-art SISO and MISO models, as well as a baseline MIMO model we constructed. Our evaluation highlights the real-world applicability of MIMONet and its potential to significantly enhance the performance of intelligent robotic systems.

Zhongnan Qu, Zimu Zhou, Yun Cheng et al.

We investigate the compression of deep neural networks by quantizing their weights and activations into multiple binary bases, known as multi-bit networks (MBNs), which accelerate the inference and reduce the storage for the deployment on low-resource mobile and embedded platforms. We propose Adaptive Loss-aware Quantization (ALQ), a new MBN quantization pipeline that is able to achieve an average bitwidth below one-bit without notable loss in inference accuracy. Unlike previous MBN quantization solutions that train a quantizer by minimizing the error to reconstruct full precision weights, ALQ directly minimizes the quantization-induced error on the loss function involving neither gradient approximation nor full precision maintenance. ALQ also exploits strategies including adaptive bitwidth, smooth bitwidth reduction, and iterative trained quantization to allow a smaller network size without loss in accuracy. Experiment results on popular image datasets show that ALQ outperforms state-of-the-art compressed networks in terms of both storage and accuracy. Code is available at https://github.com/zqu1992/ALQ

Dong Wang, Haris Šikić, Lothar Thiele et al.

We introduce model folding, a novel data-free model compression technique that merges structurally similar neurons across layers, significantly reducing the model size without the need for fine-tuning or access to training data. Unlike existing methods, model folding preserves data statistics during compression by leveraging k-means clustering, and using novel data-free techniques to prevent variance collapse or explosion. Our theoretical framework and experiments across standard benchmarks, including ResNet18 and LLaMA-7B, demonstrate that model folding achieves comparable performance to data-driven compression techniques and outperforms recently proposed data-free methods, especially at high sparsity levels. This approach is particularly effective for compressing large-scale models, making it suitable for deployment in resource-constrained environments.

Olga Saukh, Dong Wang, Haris Šikić et al.

Compressing neural networks without retraining is vital for deployment at scale. We study calibration-free compression through the lens of projection geometry: structured pruning is an axis-aligned projection, whereas model folding performs a low-rank projection via weight clustering. We formalize both as orthogonal operators and show that, within a rank distance of one, folding provably yields smaller parameter reconstruction error, and under mild smoothness assumptions, smaller functional perturbations than pruning. At scale, we evaluate >1000 checkpoints spanning ResNet18, PreActResNet18, ViT-B/32, and CLIP ViT-B/32 on CIFAR-10 and ImageNet-1K, covering diverse training hyperparameters (optimizers, learning rates, augmentations, regularization, sharpness-aware training), as well as multiple LLaMA-family 60M and 130M parameter models trained on C4. We show that folding typically achieves higher post-compression accuracy, with the largest gains at moderate-high compression. The gap narrows and occasionally reverses at specific training setups. Our results position folding as a geometry-aware, calibration-free alternative to pruning that is often superior in practice and principled in theory.

Shuo Wang, Mengfan Teng, Yun Cheng et al.

High-resolution mapping of fine particulate matter (PM2.5) is a cornerstone of sustainable urbanism but remains critically hindered by the spatial sparsity of ground monitoring networks. While traditional data-driven methods attempt to bridge this gap using satellite Aerosol Optical Depth (AOD), they often suffer from severe, non-random data missingness (e.g., due to cloud cover or nighttime) and inversion biases. To overcome these limitations, this study proposes the Spatiotemporal Physics-Guided Inference Network (SPIN), a novel framework designed for inductive spatiotemporal kriging. Unlike conventional approaches, SPIN synergistically integrates domain knowledge into deep learning by explicitly modeling physical advection and diffusion processes via parallel graph kernels. Crucially, we introduce a paradigm-shifting training strategy: rather than using error-prone AOD as a direct input, we repurpose it as a spatial gradient constraint within the loss function. This allows the model to learn structural pollution patterns from satellite data while remaining robust to data voids. Validated in the highly polluted Beijing-Tianjin-Hebei and Surrounding Areas (BTHSA), SPIN achieves a new state-of-the-art with a Mean Absolute Error (MAE) of 9.52 ug/m^3, effectively generating continuous, physically plausible pollution fields even in unmonitored areas. This work provides a robust, low-cost, and all-weather solution for fine-grained environmental management.

Dong Wang, Olga Saukh, Xiaoxi He et al.

While the deployment of deep learning models on edge devices is increasing, these models often lack robustness when faced with dynamic changes in sensed data. This can be attributed to sensor drift, or variations in the data compared to what was used during offline training due to factors such as specific sensor placement or naturally changing sensing conditions. Hence, achieving the desired robustness necessitates the utilization of either an invariant architecture or specialized training approaches, like data augmentation techniques. Alternatively, input transformations can be treated as a domain shift problem, and solved by post-deployment model adaptation. In this paper, we train a parameterized subspace of configurable networks, where an optimal network for a particular parameter setting is part of this subspace. The obtained subspace is low-dimensional and has a surprisingly simple structure even for complex, non-invertible transformations of the input, leading to an exceptionally high efficiency of subspace-configurable networks (SCNs) when limited storage and computing resources are at stake.

Wenying Duan, Xiaoxi He, Lu Zhou et al.

Time series forecasting has widespread applications in urban life ranging from air quality monitoring to traffic analysis. However, accurate time series forecasting is challenging because real-world time series suffer from the distribution shift problem, where their statistical properties change over time. Despite extensive solutions to distribution shifts in domain adaptation or generalization, they fail to function effectively in unknown, constantly-changing distribution shifts, which are common in time series. In this paper, we propose Hyper Time- Series Forecasting (HTSF), a hypernetwork-based framework for accurate time series forecasting under distribution shift. HTSF jointly learns the time-varying distributions and the corresponding forecasting models in an end-to-end fashion. Specifically, HTSF exploits the hyper layers to learn the best characterization of the distribution shifts, generating the model parameters for the main layers to make accurate predictions. We implement HTSF as an extensible framework that can incorporate diverse time series forecasting models such as RNNs and Transformers. Extensive experiments on 9 benchmarks demonstrate that HTSF achieves state-of-the-art performances.

Andres Gomez, Andreas Tretter, Pascal Alexander Hager et al.

Sensing systems powered by energy harvesting have traditionally been designed to tolerate long periods without energy. As the Internet of Things (IoT) evolves towards a more transient and opportunistic execution paradigm, reducing energy storage costs will be key for its economic and ecologic viability. However, decreasing energy storage in harvesting systems introduces reliability issues. Transducers only produce intermittent energy at low voltage and current levels, making guaranteed task completion a challenge. Existing ad hoc methods overcome this by buffering enough energy either for single tasks, incurring large data-retention overheads, or for one full application cycle, requiring a large energy buffer. We present Julienning: an automated method for optimizing the total energy cost of batteryless applications. Using a custom specification model, developers can describe transient applications as a set of atomically executed kernels with explicit data dependencies. Our optimization flow can partition data- and energy-intensive applications into multiple execution cycles with bounded energy consumption. By leveraging interkernel data dependencies, these energy-bounded execution cycles minimize the number of system activations and nonvolatile data transfers, and thus the total energy overhead. We validate our methodology with two batteryless cameras running energy-intensive machine learning applications. Results demonstrate that compared to ad hoc solutions, our method can reduce the required energy storage by over 94% while only incurring a 0.12% energy overhead.

Matthias Meyer, Michaela Wenner, Clément Hibert et al.

Real-world datasets collected with sensor networks often contain incomplete and uncertain labels as well as artefacts arising from the system environment. Complete and reliable labeling is often infeasible for large-scale and long-term sensor network deployments due to the labor and time overhead, limited availability of experts and missing ground truth. In addition, if the machine learning method used for analysis is sensitive to certain features of a deployment, labeling and learning needs to be repeated for every new deployment. To address these challenges, we propose to make use of system context information formalized in an information graph and embed it in the learning process via contrastive learning. Based on real-world data we show that this approach leads to an increased accuracy in case of weakly labeled data and leads to an increased robustness and transferability of the classifier to new sensor locations.

Lennart Heim, Andreas Biri, Zhongnan Qu et al.

With the surge of inexpensive computational and memory resources, neural networks (NNs) have experienced an unprecedented growth in architectural and computational complexity. Introducing NNs to resource-constrained devices enables cost-efficient deployments, widespread availability, and the preservation of sensitive data. This work addresses the challenges of bringing Machine Learning to MCUs, where we focus on the ubiquitous ARM Cortex-M architecture. The detailed effects and trade-offs that optimization methods, software frameworks, and MCU hardware architecture have on key performance metrics such as inference latency and energy consumption have not been previously studied in depth for state-of-the-art frameworks such as TensorFlow Lite Micro. We find that empirical investigations which measure the perceptible metrics - performance as experienced by the user - are indispensable, as the impact of specialized instructions and layer types can be subtle. To this end, we propose an implementation-aware design as a cost-effective method for verification and benchmarking. Employing our developed toolchain, we demonstrate how existing NN deployments on resource-constrained devices can be improved by systematically optimizing NNs to their targeted application scenario.

Zhongnan Qu, Cong Liu, Lothar Thiele

Emerging edge intelligence applications require the server to retrain and update deep neural networks deployed on remote edge nodes to leverage newly collected data samples. Unfortunately, it may be impossible in practice to continuously send fully updated weights to these edge nodes due to the highly constrained communication resource. In this paper, we propose the weight-wise deep partial updating paradigm, which smartly selects a small subset of weights to update in each server-to-edge communication round, while achieving a similar performance compared to full updating. Our method is established through analytically upper-bounding the loss difference between partial updating and full updating, and only updates the weights which make the largest contributions to the upper bound. Extensive experimental results demonstrate the efficacy of our partial updating methodology which achieves a high inference accuracy while updating a rather small number of weights.

Xiaoxi He, Dawei Gao, Zimu Zhou et al.

Many mobile applications demand selective execution of multiple correlated deep learning inference tasks on resource-constrained platforms. Given a set of deep neural networks, each pre-trained for a single task, it is desired that executing arbitrary combinations of tasks yields minimal computation cost. Pruning each network separately yields suboptimal computation cost due to task relatedness. A promising remedy is to merge the networks into a multitask network to eliminate redundancy across tasks before network pruning. However, pruning a multitask network combined by existing network merging schemes cannot minimise the computation cost of every task combination because they do not consider such a future pruning. To this end, we theoretically identify the conditions such that pruning a multitask network minimises the computation of all task combinations. On this basis, we propose Pruning-Aware Merging (PAM), a heuristic network merging scheme to construct a multitask network that approximates these conditions. The merged network is then ready to be further pruned by existing network pruning methods. Evaluations with different pruning schemes, datasets, and network architectures show that PAM achieves up to 4.87x less computation against the baseline without network merging, and up to 2.01x less computation against the baseline with a state-of-the-art network merging scheme.

Matthias Meyer, Timo Farei-Campagna, Akos Pasztor et al.

In natural hazard warning systems fast decision making is vital to avoid catastrophes. Decision making at the edge of a wireless sensor network promises fast response times but is limited by the availability of energy, data transfer speed, processing and memory constraints. In this work we present a realization of a wireless sensor network for hazard monitoring based on an array of event-triggered single-channel micro-seismic sensors with advanced signal processing and characterization capabilities based on a novel co-detection technique. On the one hand we leverage an ultra-low power, threshold-triggering circuit paired with on-demand digital signal acquisition capable of extracting relevant information exactly and efficiently at times when it matters most and consequentially not wasting precious resources when nothing can be observed. On the other hand we utilize machine-learning-based classification implemented on low-power, off-the-shelf microcontrollers to avoid false positive warnings and to actively identify humans in hazard zones. The sensors' response time and memory requirement is substantially improved by quantizing and pipelining the inference of a convolutional neural network. In this way, convolutional neural networks that would not run unmodified on a memory constrained device can be executed in real-time and at scale on low-power embedded devices. A field study with our system is running on the rockfall scarp of the Matterhorn Hörnligrat at 3500 m a.s.l. since 08/2018.

Xiaoxi He, Zimu Zhou, Lothar Thiele

Future mobile devices are anticipated to perceive, understand and react to the world on their own by running multiple correlated deep neural networks on-device. Yet the complexity of these neural networks needs to be trimmed down both within-model and cross-model to fit in mobile storage and memory. Previous studies focus on squeezing the redundancy within a single neural network. In this work, we aim to reduce the redundancy across multiple models. We propose Multi-Task Zipping (MTZ), a framework to automatically merge correlated, pre-trained deep neural networks for cross-model compression. Central in MTZ is a layer-wise neuron sharing and incoming weight updating scheme that induces a minimal change in the error function. MTZ inherits information from each model and demands light retraining to re-boost the accuracy of individual tasks. Evaluations show that MTZ is able to fully merge the hidden layers of two VGG-16 networks with a 3.18% increase in the test error averaged on ImageNet and CelebA, or share 39.61% parameters between the two networks with <0.5% increase in the test errors for both tasks. The number of iterations to retrain the combined network is at least 17.8 times lower than that of training a single VGG-16 network. Moreover, experiments show that MTZ is also able to effectively merge multiple residual networks.

Matthias Meyer, Jan Beutel, Lothar Thiele

Identifying acoustic events from a continuously streaming audio source is of interest for many applications including environmental monitoring for basic research. In this scenario neither different event classes are known nor what distinguishes one class from another. Therefore, an unsupervised feature learning method for exploration of audio data is presented in this paper. It incorporates the two following novel contributions: First, an audio frame predictor based on a Convolutional LSTM autoencoder is demonstrated, which is used for unsupervised feature extraction. Second, a training method for autoencoders is presented, which leads to distinct features by amplifying event similarities. In comparison to standard approaches, the features extracted from the audio frame predictor trained with the novel approach show 13 % better results when used with a classifier and 36 % better results when used for clustering.

Matthias Meyer, Lukas Cavigelli, Lothar Thiele

Wireless distributed systems as used in sensor networks, Internet-of-Things and cyber-physical systems, impose high requirements on resource efficiency. Advanced preprocessing and classification of data at the network edge can help to decrease the communication demand and to reduce the amount of data to be processed centrally. In the area of distributed acoustic sensing, the combination of algorithms with a high classification rate and resource-constraint embedded systems is essential. Unfortunately, algorithms for acoustic event detection have a high memory and computational demand and are not suited for execution at the network edge. This paper addresses these aspects by applying structural optimizations to a convolutional neural network for audio event detection to reduce the memory requirement by a factor of more than 500 and the computational effort by a factor of 2.1 while performing 9.2% better.

Ramya Jayaram Masti, Devendra Rai, Aanjhan Ranganathan et al.

Side channels remain a challenge to information flow control and security in modern computing platforms. Resource partitioning techniques that minimise the number of shared resources among processes are often used to address this challenge. In this work, we focus on multi-core platforms and we demonstrate that even seemingly strong isolation techniques based on dedicated cores and memory can be circumvented through the use of thermal side channels. Specifically, we show that the processor core temperature can be used both as a side channel as well as a covert communication channel even when the system implements strong spatial and temporal partitioning. Our experiments on an x86-based platform demonstrate covert thermal channels that achieve up to 12.5 bps and a weak side channel that can detect processes executed on neighbouring cores. This work therefore shows a limitation in the isolation that can be achieved on existing multi-core systems.

Pier Stanislao Paolucci, Iuliana Bacivarov, Devendra Rai et al.

The EURETILE project required the selection and coding of a set of dedicated benchmarks. The project is about the software and hardware architecture of future many-tile distributed fault-tolerant systems. We focus on dynamic workloads characterised by heavy numerical processing requirements. The ambition is to identify common techniques that could be applied to both the Embedded Systems and HPC domains. This document is the first public deliverable of Work Package 7: Challenging Tiled Applications.

Pier Stanislao Paolucci, Iuliana Bacivarov, Gert Goossens et al.

This is the summary of first three years of activity of the EURETILE FP7 project 247846. EURETILE investigates and implements brain-inspired and fault-tolerant foundational innovations to the system architecture of massively parallel tiled computer architectures and the corresponding programming paradigm. The execution targets are a many-tile HW platform, and a many-tile simulator. A set of SW process - HW tile mapping candidates is generated by the holistic SW tool-chain using a combination of analytic and bio-inspired methods. The Hardware dependent Software is then generated, providing OS services with maximum efficiency/minimal overhead. The many-tile simulator collects profiling data, closing the loop of the SW tool chain. Fine-grain parallelism inside processes is exploited by optimized intra-tile compilation techniques, but the project focus is above the level of the elementary tile. The elementary HW tile is a multi-processor, which includes a fault tolerant Distributed Network Processor (for inter-tile communication) and ASIP accelerators. Furthermore, EURETILE investigates and implements the innovations for equipping the elementary HW tile with high-bandwidth, low-latency brain-like inter-tile communication emulating 3 levels of connection hierarchy, namely neural columns, cortical areas and cortex, and develops a dedicated cortical simulation benchmark: DPSNN-STDP (Distributed Polychronous Spiking Neural Net with synaptic Spiking Time Dependent Plasticity). EURETILE leverages on the multi-tile HW paradigm and SW tool-chain developed by the FET-ACA SHAPES Integrated Project (2006-2009).