arXiv subjects

Manuel Roveri

Publications and source records attributed to Manuel Roveri.

At least 19 recordsLinked to original sources

arXiv2026

What changes after deployment? A survey on On-device Learning in TinyML

Machine learning models on microcontroller-class devices (TinyML) face a fundamental challenge: post-deployment distribution change undermines static models. On-device learning (ODL) addresses this by running the learning process directly on the device. The existing literature has not characterized how distribution change occurs or how different change types require different solutions. Approximately 70 ODL works are surveyed under one principle: the distribution change regime. The survey analyzes how different types of distribution change influence the applications addressable on-device, the hardware employed, and the structure of the solutions. A persistent gap between methodological benchmarks and real-world deployment scenarios is also identified.

cs.LG↗

arXiv2026

HERCULES: Hardware-Efficient, Robust, Continual Learning Neural Architecture Search

Neural Architecture Search (NAS) has emerged as a powerful framework for automatically discovering neural architectures that balance accuracy and efficiency. However, as AI transitions from static benchmarks to real-world deployment, the traditional focus on hardware-aware efficiency is no longer sufficient. We observe that modern NAS methods, especially those that target edge AI, are evolving to address a triple objective: Efficiency, Robustness, and Continual Learning. While efficiency ensures feasibility in resource-constrained environments, robustness guarantees reliability under environmental variabilities, and continual learning enables adaptation to sequential tasks without catastrophic forgetting. We propose a taxonomy of NAS approaches through this triple lens, distinguishing between methods targeting resource optimization, environmental resilience, and architectural plasticity. This unified perspective reveals that these axes, though often studied in isolation, are mutually reinforcing. Building on this taxonomy, we map the current landscape of these NAS methods into a new framework called Hardware-Efficient, Robust, and ContinUal LEarning Search (HERCULES). We define the desiderata, the twelve labours of HERCULES, addressing the non-trivial challenge of balancing an adequate search-space exploration with the immense computational costs of a multi-objective NAS, accounting for these crucial objectives of current AI systems. By identifying critical gaps in existing research, this survey outlines a roadmap toward integrated algorithmic, architectural, and hardware-software co-design for truly deployable, lifelong-learning AI systems.

cs.LG↗

arXiv2026

Position Paper: From Edge AI to Adaptive Edge AI

Edge AI is often framed as model compression and deployment under tight constraints. We argue a stronger operational thesis: Edge AI in realistic deployments is necessarily adaptive. In long-horizon operation, a fixed (non-adaptive) configuration faces a fundamental failure mode: as data and operating conditions evolve and change in time, it must either (i) violate time-varying budgets (latency/energy/thermal/connectivity/privacy) or (ii) lose predictive reliability (accuracy and, critically, calibration), with risk concentrating in transient regimes and rare time intervals rather than in average performance. If a deployed system cannot reconfigure its computation - and, when required, its model state - under evolving conditions and constraints, it reduces to static embedded inference and cannot provide sustained utility. This position paper introduces a minimal Agent-System-Environment (ASE) lens that makes adaptivity precise at the edge by specifying (i) what changes, (ii) what is observed, (iii) what can be reconfigured, and (iv) which constraints must remain satisfied over time. Building on this framing, we formulate ten research challenges for the next decade, spanning theoretical guarantees for evolving systems, dynamic architectures and hybrid transitions between data-driven and model-based components, fault/anomaly-driven targeted updates, System-1/System-2 decompositions (anytime intelligence), modularity, validation under scarce labels, and evaluation protocols that quantify lifecycle efficiency and recovery/stability under drift and interventions.

cs.AR↗

arXiv2026

SQUAD: Scalable Quorum Adaptive Decisions via ensemble of early exit neural networks

Early-exit neural networks have become popular for reducing inference latency by allowing intermediate predictions when sufficient confidence is achieved. However, standard approaches typically rely on single-model confidence thresholds, which are frequently unreliable due to inherent calibration issues. To address this, we introduce SQUAD (Scalable Quorum Adaptive Decisions), the first inference scheme that integrates early-exit mechanisms with distributed ensemble learning, improving uncertainty estimation while reducing the inference time. Unlike traditional methods that depend on individual confidence scores, SQUAD employs a quorum-based stopping criterion on early-exit learners by collecting intermediate predictions incrementally in order of computational complexity until a consensus is reached and halting the computation at that exit if the consensus is statistically significant. To maximize the efficacy of this voting mechanism, we also introduce QUEST (Quorum Search Technique), a Neural Architecture Search method to select early-exit learners with optimized hierarchical diversity, ensuring learners are complementary at every intermediate layer. This consensus-driven approach yields statistically robust early exits, improving the test accuracy up to 5.95% compared to state-of-the-art dynamic solutions with a comparable computational cost and reducing the inference latency up to 70.60% compared to static ensembles while maintaining a good accuracy.

cs.LG↗

arXiv2025

BEP: A Binary Error Propagation Algorithm for Binary Neural Networks Training

Binary Neural Networks (BNNs), which constrain both weights and activations to binary values, offer substantial reductions in computational complexity, memory footprint, and energy consumption. These advantages make them particularly well suited for deployment on resource-constrained devices. However, training BNNs via gradient-based optimization remains challenging due to the discrete nature of their variables. The dominant approach, quantization-aware training, circumvents this issue by employing surrogate gradients. Yet, this method requires maintaining latent full-precision parameters and performing the backward pass with floating-point arithmetic, thereby forfeiting the efficiency of binary operations during training. While alternative approaches based on local learning rules exist, they are unsuitable for global credit assignment and for back-propagating errors in multi-layer architectures. This paper introduces Binary Error Propagation (BEP), the first learning algorithm to establish a principled, discrete analog of the backpropagation chain rule. This mechanism enables error signals, represented as binary vectors, to be propagated backward through multiple layers of a neural network. BEP operates entirely on binary variables, with all forward and backward computations performed using only bitwise operations. Crucially, this makes BEP the first solution to enable end-to-end binary training for recurrent neural network architectures. We validate the effectiveness of BEP on both multi-layer perceptrons and recurrent neural networks, demonstrating gains of up to +6.89% and +10.57% in test accuracy, respectively. The proposed algorithm is released as an open-source repository.

cs.LG↗

arXiv2025

DQT: Dynamic Quantization Training via Dequantization-Free Nested Integer Arithmetic

The deployment of deep neural networks on resource-constrained devices relies on quantization. While static, uniform quantization applies a fixed bit-width to all inputs, it fails to adapt to their varying complexity. Dynamic, instance-based mixed-precision quantization promises a superior accuracy-efficiency trade-off by allocating higher precision only when needed. However, a critical bottleneck remains: existing methods require a costly dequantize-to-float and requantize-to-integer cycle to change precision, breaking the integer-only hardware paradigm and compromising performance gains. This paper introduces Dynamic Quantization Training (DQT), a novel framework that removes this bottleneck. At the core of DQT is a nested integer representation where lower-precision values are bit-wise embedded within higher-precision ones. This design, coupled with custom integer-only arithmetic, allows for on-the-fly bit-width switching through a near-zero-cost bit-shift operation. This makes DQT the first quantization framework to enable both dequantization-free static mixed-precision of the backbone network, and truly efficient dynamic, instance-based quantization through a lightweight controller that decides at runtime how to quantize each layer. We demonstrate DQT state-of-the-art performance on ResNet18 on CIFAR-10 and ResNet50 on ImageNet. On ImageNet, our 4-bit dynamic ResNet50 achieves 77.00% top-1 accuracy, an improvement over leading static (LSQ, 76.70%) and dynamic (DQNET, 76.94%) methods at a comparable BitOPs budget. Crucially, DQT achieves this with a bit-width transition cost of only 28.3M simple bit-shift operations, a drastic improvement over the 56.6M costly Multiply-Accumulate (MAC) floating-point operations required by previous dynamic approaches - unlocking a new frontier in efficient, adaptive AI.

cs.LG↗

arXiv2025

InfoQ: Mixed-Precision Quantization via Global Information Flow

Mixed-precision quantization (MPQ) is crucial for deploying deep neural networks on resource-constrained devices, but finding the optimal bit-width for each layer represents a complex combinatorial optimization problem. Current state-of-the-art methods rely on computationally expensive search algorithms or local sensitivity heuristic proxies like the Hessian, which fail to capture the cascading global effects of quantization error. In this work, we argue that the quantization sensitivity of a layer should not be measured by its local properties, but by its impact on the information flow throughout the entire network. We introduce InfoQ, a novel framework for MPQ that is training-free in the bit-width search phase. InfoQ assesses layer sensitivity by quantizing each layer at different bit-widths and measuring, through a single forward pass, the resulting change in mutual information in the subsequent layers. This quantifies how much each layer quantization impacts the network information flow. The resulting scores are used to formulate bit-width allocation as an integer linear programming problem, which is solved efficiently to minimize total sensitivity under a given budget (e.g., model size or BitOps). Our retraining-free search phase provides a superior search-time/accuracy trade-off (using two orders of magnitude less data compared to state-of-the-art methods such as LIMPQ), while yielding up to a 1% accuracy improvement for MobileNetV2 and ResNet18 on ImageNet at high compression rates (14X and 10.66X).

cs.LG↗

arXiv2025

SEAL: Searching Expandable Architectures for Incremental Learning

Incremental learning is a machine learning paradigm where a model learns from a sequential stream of tasks. This setting poses a key challenge: balancing plasticity (learning new tasks) and stability (preserving past knowledge). Neural Architecture Search (NAS), a branch of AutoML, automates the design of the architecture of Deep Neural Networks and has shown success in static settings. However, existing NAS-based approaches to incremental learning often rely on expanding the model at every task, making them impractical in resource-constrained environments. In this work, we introduce SEAL, a NAS-based framework tailored for data-incremental learning, a scenario where disjoint data samples arrive sequentially and are not stored for future access. SEAL adapts the model structure dynamically by expanding it only when necessary, based on a capacity estimation metric. Stability is preserved through cross-distillation training after each expansion step. The NAS component jointly searches for both the architecture and the optimal expansion policy. Experiments across multiple benchmarks demonstrate that SEAL effectively reduces forgetting and enhances accuracy while allocating additional capacity only when required. These results highlight the promise of combining NAS and selective expansion for efficient, adaptive learning in incremental scenarios.

cs.LG↗

arXiv2025

TActiLE: Tiny Active LEarning for wearable devices

Tiny Machine Learning (TinyML) algorithms have seen extensive use in recent years, enabling wearable devices to be not only connected but also genuinely intelligent by running machine learning (ML) computations directly on-device. Among such devices, smart glasses have particularly benefited from TinyML advancements. TinyML facilitates the on-device execution of the inference phase of ML algorithms on embedded and wearable devices, and more recently, it has expanded into On-device Learning (ODL), which allows both inference and learning phases to occur directly on the device. The application of ODL techniques to wearable devices is particularly compelling, as it enables the development of more personalized models that adapt based on the data of the user. However, one of the major challenges of ODL algorithms is the scarcity of labeled data collected on-device. In smart wearable contexts, requiring users to manually label large amounts of data is often impractical and could lead to user disengagement with the technology. To address this issue, this paper explores the application of Active Learning (AL) techniques, i.e., techniques that aim at minimizing the labeling effort, by actively selecting from a large quantity of unlabeled data only a small subset to be labeled and added to the training set of the algorithm. In particular, we propose TActiLE, a novel AL algorithm that selects from the stream of on-device sensor data the ones that would help the ML algorithm improve the most once coupled with labels provided by the user. TActiLE is the first Active Learning technique specifically designed for the TinyML context. We evaluate its effectiveness and efficiency through experiments on multiple image classification datasets. The results demonstrate its suitability for tiny and wearable devices.

cs.LG↗

arXiv2025

DYNAMAX: Dynamic computing for Transformers and Mamba based architectures

Early exits (EEs) offer a promising approach to reducing computational costs and latency by dynamically terminating inference once a satisfactory prediction confidence on a data sample is achieved. Although many works integrate EEs into encoder-only Transformers, their application to decoder-only architectures and, more importantly, Mamba models, a novel family of state-space architectures in the LLM realm, remains insufficiently explored. This work introduces DYNAMAX, the first framework to exploit the unique properties of Mamba architectures for early exit mechanisms. We not only integrate EEs into Mamba but also repurpose Mamba as an efficient EE classifier for both Mamba-based and transformer-based LLMs, showcasing its versatility. Our experiments employ the Mistral 7B transformer compared to the Codestral 7B Mamba model, using data sets such as TruthfulQA, CoQA, and TriviaQA to evaluate computational savings, accuracy, and consistency. The results highlight the adaptability of Mamba as a powerful EE classifier and its efficiency in balancing computational cost and performance quality across NLP tasks. By leveraging Mamba's inherent design for dynamic processing, we open pathways for scalable and efficient inference in embedded applications and resource-constrained environments. This study underscores the transformative potential of Mamba in redefining dynamic computing paradigms for LLMs.

cs.CL↗

arXiv2025

On-Sensor Convolutional Neural Networks with Early-Exits

Tiny Machine Learning (TinyML) is a novel research field aiming at integrating Machine Learning (ML) within embedded devices with limited memory, computation, and energy. Recently, a new branch of TinyML has emerged, focusing on integrating ML directly into the sensors to further reduce the power consumption of embedded devices. Interestingly, despite their state-of-the-art performance in many tasks, none of the current solutions in the literature aims to optimize the implementation of Convolutional Neural Networks (CNNs) operating directly into sensors. In this paper, we introduce for the first time in the literature the optimized design and implementation of Depth-First CNNs operating on the Intelligent Sensor Processing Unit (ISPU) within an Inertial Measurement Unit (IMU) by STMicroelectronics. Our approach partitions the CNN between the ISPU and the microcontroller (MCU) and employs an Early-Exit mechanism to stop the computations on the IMU when enough confidence about the results is achieved, hence significantly reducing power consumption. When using a NUCLEO-F411RE board, this solution achieved an average current consumption of 4.8 mA, marking an 11% reduction compared to the regular inference pipeline on the MCU, while having equal accuracy.

cs.LG↗

arXiv2025

Architecture-Aware Minimization (A$^2$M): How to Find Flat Minima in Neural Architecture Search

Neural Architecture Search (NAS) has become an essential tool for designing effective and efficient neural networks. In this paper, we investigate the geometric properties of neural architecture spaces commonly used in differentiable NAS methods, specifically NAS-Bench-201 and DARTS. By defining flatness metrics such as neighborhoods and loss barriers along paths in architecture space, we reveal locality and flatness characteristics analogous to the well-known properties of neural network loss landscapes in weight space. In particular, we find that highly accurate architectures cluster together in flat regions, while suboptimal architectures remain isolated, unveiling the detailed geometrical structure of the architecture search landscape. Building on these insights, we propose Architecture-Aware Minimization (A$^2$M), a novel analytically derived algorithmic framework that explicitly biases, for the first time, the gradient of differentiable NAS methods towards flat minima in architecture space. A$^2$M consistently improves generalization over state-of-the-art DARTS-based algorithms on benchmark datasets including CIFAR-10, CIFAR-100, and ImageNet16-120, across both NAS-Bench-201 and DARTS search spaces. Notably, A$^2$M is able to increase the test accuracy, on average across different differentiable NAS methods, by +3.60\% on CIFAR-10, +4.60\% on CIFAR-100, and +3.64\% on ImageNet16-120, demonstrating its superior effectiveness in practice. A$^2$M can be easily integrated into existing differentiable NAS frameworks, offering a versatile tool for future research and applications in automated machine learning. We open-source our code at https://github.com/AI-Tech-Research-Lab/AsquaredM.

cs.LG↗

arXiv2025

Dendron: Enhancing Human Activity Recognition with On-Device TinyML Learning

Human activity recognition (HAR) is a research field that employs Machine Learning (ML) techniques to identify user activities. Recent studies have prioritized the development of HAR solutions directly executed on wearable devices, enabling the on-device activity recognition. This approach is supported by the Tiny Machine Learning (TinyML) paradigm, which integrates ML within embedded devices with limited resources. However, existing approaches in the field lack in the capability for on-device learning of new HAR tasks, particularly when supervised data are scarce. To address this limitation, our paper introduces Dendron, a novel TinyML methodology designed to facilitate the on-device learning of new tasks for HAR, even in conditions of limited supervised data. Experimental results on two public-available datasets and an off-the-shelf device (STM32-NUCLEO-F401RE) show the effectiveness and efficiency of the proposed solution.

cs.LG↗

arXiv2025

EmbBERT: Attention Under 2 MB Memory

Transformer architectures based on the attention mechanism have revolutionized natural language processing (NLP), driving major breakthroughs across virtually every NLP task. However, their substantial memory and computational requirements still hinder deployment on ultra-constrained devices such as wearables and Internet-of-Things (IoT) units, where available memory is limited to just a few megabytes. To address this challenge, we introduce EmbBERT, a tiny language model (TLM) architecturally designed for extreme efficiency. The model integrates a compact embedding layer, streamlined feed-forward blocks, and an efficient attention mechanism that together enable optimal performance under strict memory budgets. Through this redesign for the extreme edge, we demonstrate that highly simplified transformer architectures remain remarkably effective under tight resource constraints. EmbBERT requires only 2 MB of total memory, and achieves accuracy performance comparable to the ones of state-of-the-art (SotA) models that require a $\mathbf{10\times}$ memory budget. Extensive experiments on the curated TinyNLP benchmark and the GLUE suite confirm that EmbBERT achieves competitive accuracy, comparable to that of larger SotA models, and consistently outperforms downsized versions of BERT and MAMBA of similar size. Furthermore, we demonstrate the model resilience to 8-bit quantization, which further reduces memory usage to just 781 kB , and the scalability of the EmbBERT architecture across the sub-megabyte to tens-of-megabytes range. Finally, we perform an ablation study demonstrating the positive contributions of all components and the pre-training procedure. All code, scripts, and checkpoints are publicly released to ensure reproducibility: https://github.com/RiccardoBravin/tiny-LLM.

cs.CL↗

arXiv2025

Quantifying Cryptocurrency Unpredictability: A Comprehensive Study of Complexity and Forecasting

This paper offers a thorough examination of the univariate predictability in cryptocurrency time-series. By exploiting a combination of complexity measure and model predictions we explore the cryptocurrencies time-series forecasting task focusing on the exchange rate in USD of Litecoin, Binance Coin, Bitcoin, Ethereum, and XRP. On one hand, to assess the complexity and the randomness of these time-series, a comparative analysis has been performed using Brownian and colored noises as a benchmark. The results obtained from the Complexity-Entropy causality plane and power density spectrum analysis reveal that cryptocurrency time-series exhibit characteristics closely resembling those of Brownian noise when analyzed in a univariate context. On the other hand, the application of a wide range of statistical, machine and deep learning models for time-series forecasting demonstrates the low predictability of cryptocurrencies. Notably, our analysis reveals that simpler models such as Naive models consistently outperform the more complex machine and deep learning ones in terms of forecasting accuracy across different forecast horizons and time windows. The combined study of complexity and forecasting accuracies highlights the difficulty of predicting the cryptocurrency market. These findings provide valuable insights into the inherent characteristics of the cryptocurrency data and highlight the need to reassess the challenges associated with predicting cryptocurrency's price movements.

q-fin.ST↗

arXiv2024

Training Multi-Layer Binary Neural Networks With Local Binary Error Signals

Binary Neural Networks (BNNs) significantly reduce computational complexity and memory usage in machine and deep learning by representing weights and activations with just one bit. However, most existing training algorithms for BNNs rely on quantization-aware floating-point Stochastic Gradient Descent (SGD), limiting the full exploitation of binary operations to the inference phase only. In this work, we propose, for the first time, a fully binary and gradient-free training algorithm for multi-layer BNNs, eliminating the need for back-propagated floating-point gradients. Specifically, the proposed algorithm relies on local binary error signals and binary weight updates, employing integer-valued hidden weights that serve as a synaptic metaplasticity mechanism, thereby enhancing its neurobiological plausibility. Our proposed solution enables the training of binary multi-layer perceptrons by using exclusively XNOR, Popcount, and increment/decrement operations. Experimental results on multi-class classification benchmarks show test accuracy improvements of up to +35.47% over the only existing fully binary single-layer state-of-the-art solution. Compared to full-precision SGD, our solution improves test accuracy by up to +35.30% under the same total memory demand, while also reducing computational cost by two to three orders of magnitude in terms of the total number of Boolean gates. The proposed algorithm is made available to the scientific community as a public repository.

cs.LG↗

arXiv2024

TIFeD: a Tiny Integer-based Federated learning algorithm with Direct feedback alignment

Training machine and deep learning models directly on extremely resource-constrained devices is the next challenge in the field of tiny machine learning. The related literature in this field is very limited, since most of the solutions focus only on on-device inference or model adaptation through online learning, leaving the training to be carried out on external Cloud services. An interesting technological perspective is to exploit Federated Learning (FL), which allows multiple devices to collaboratively train a shared model in a distributed way. However, the main drawback of state-of-the-art FL algorithms is that they are not suitable for running on tiny devices. For the first time in the literature, in this paper we introduce TIFeD, a Tiny Integer-based Federated learning algorithm with Direct Feedback Alignment (DFA) entirely implemented by using an integer-only arithmetic and being specifically designed to operate on devices with limited resources in terms of memory, computation and energy. Besides the traditional full-network operating modality, in which each device of the FL setting trains the entire neural network on its own local data, we propose an innovative single-layer TIFeD implementation, which enables each device to train only a portion of the neural network model and opens the door to a new way of distributing the learning procedure across multiple devices. The experimental results show the feasibility and effectiveness of the proposed solution. The proposed TIFeD algorithm, with its full-network and single-layer implementations, is made available to the scientific community as a public repository.

cs.LG↗

arXiv2024

StreamTinyNet: video streaming analysis with spatial-temporal TinyML

Tiny Machine Learning (TinyML) is a branch of Machine Learning (ML) that constitutes a bridge between the ML world and the embedded system ecosystem (i.e., Internet of Things devices, embedded devices, and edge computing units), enabling the execution of ML algorithms on devices constrained in terms of memory, computational capabilities, and power consumption. Video Streaming Analysis (VSA), one of the most interesting tasks of TinyML, consists in scanning a sequence of frames in a streaming manner, with the goal of identifying interesting patterns. Given the strict constraints of these tiny devices, all the current solutions rely on performing a frame-by-frame analysis, hence not exploiting the temporal component in the stream of data. In this paper, we present StreamTinyNet, the first TinyML architecture to perform multiple-frame VSA, enabling a variety of use cases that requires spatial-temporal analysis that were previously impossible to be carried out at a TinyML level. Experimental results on public-available datasets show the effectiveness and efficiency of the proposed solution. Finally, StreamTinyNet has been ported and tested on the Arduino Nicla Vision, showing the feasibility of what proposed.

cs.CV↗