Manuel Roveri

Luca Colombo, Fabrizio Pittorino, Daniele Zambon et al.

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.

Kleanthis Malialis, Manuel Roveri, Cesare Alippi et al.

In real-world applications, the process generating the data might suffer from nonstationary effects (e.g., due to seasonality, faults affecting sensors or actuators, and changes in the users' behaviour). These changes, often called concept drift, might induce severe (potentially catastrophic) impacts on trained learning models that become obsolete over time, and inadequate to solve the task at hand. Learning in presence of concept drift aims at designing machine and deep learning models that are able to track and adapt to concept drift. Typically, techniques to handle concept drift are either active or passive, and traditionally, these have been considered to be mutually exclusive. Active techniques use an explicit drift detection mechanism, and re-train the learning algorithm when concept drift is detected. Passive techniques use an implicit method to deal with drift, and continually update the model using incremental learning. Differently from what present in the literature, we propose a hybrid alternative which merges the two approaches, hence, leveraging on their advantages. The proposed method called Hybrid-Adaptive REBAlancing (HAREBA) significantly outperforms strong baselines and state-of-the-art methods in terms of learning quality and speed; we experiment how it is effective under severe class imbalance levels too.

Massimo Pavan, Luca Pezzarossa, Fabrizio Pittorino et al.

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.

Alberto Pirillo, Luca Colombo, Manuel Roveri

Quantization is a pivotal technique for managing the growing computational and memory demands of Deep Neural Networks (DNNs). By reducing the number of bits used to represent weights and activations (typically from 32-bit Floating-Point (FP) to 16-bit or 8-bit integers), quantization reduces memory footprint, energy consumption, and execution time of DNNs. However, most existing methods typically target DNN inference, while training still relies on FP operations, limiting applicability in environments where FP arithmetic is unavailable. To date, only one prior work has addressed integer-only training, and only for Multi-Layer Perceptron (MLP) architectures. This paper introduces NITRO-D, a novel framework for training deep integer-only Convolutional Neural Networks (CNNs) that operate entirely in the integer domain for both training and inference. NITRO-D enables training of integer CNNs without requiring a separate quantization scheme. Specifically, it introduces a novel architecture that integrates multiple local-loss blocks, which include the proposed NITRO-Scaling layer and NITRO-ReLU activation function. The proposed framework also features a novel learning algorithm that employs local error signals and leverages IntegerSGD, an optimizer specifically designed for integer computations. NITRO-D is implemented as an open-source Python library. Extensive evaluations on state-of-the-art image recognition datasets demonstrate its effectiveness. For integer-only MLPs, NITRO-D improves test accuracy by up to +5.96% over the state-of-the-art. It also successfully trains integer-only CNNs, reducing memory requirements and energy consumption by up to 76.14% and 32.42%, respectively, compared to the traditional FP backpropagation algorithm.

Matteo Gambella, Fabrizio Pittorino, Manuel Roveri

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.

Hazem Hesham Yousef Shalby, Massimo Pavan, Manuel Roveri

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.

Matteo Gambella, Fabrizio Pittorino, Giuliano Casale et al.

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.

Matteo Gambella, Fabrizio Pittorino, Manuel Roveri

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.

Riccardo Bravin, Massimo Pavan, Hazem Hesham Yousef Shalby et al.

Large Language Models (LLMs) have revolutionized natural language processing, setting new standards across a wide range of applications. However, their relevant memory and computational demands make them impractical for deployment on technologically-constrained tiny devices such as wearable devices and Internet-of-Things units. To address this limitation, we introduce EmbBERT-Q, a novel tiny language model specifically designed for tiny devices with stringent memory constraints. EmbBERT-Q achieves state-of-the-art (SotA) accuracy in Natural Language Processing tasks in this scenario, with a total memory footprint (weights and activations) of just 781 kB, representing a 25x reduction in size with respect to SotA models. By combining architectural innovations with hardware-compatible 8-bit quantization, EmbBERT-Q consistently outperforms several baseline models scaled down to a 2 MB memory budget (i.e., the maximum memory typically available in tiny devices), including heavily compressed versions of BERT and MAMBA. Extensive experimental evaluations on both a selected benchmark dataset, TinyNLP, specifically curated to evaluate Tiny Language Models in NLP tasks and real-world scenarios, and the GLUE benchmark, demonstrate EmbBERT-Q ability to deliver competitive accuracy with respect to existing approaches, achieving an unmatched balance between memory and performance. To ensure the complete and immediate reproducibility of all our results, we release all code, scripts, and model checkpoints at https://github.com/RiccardoBravin/tiny-LLM.

Matteo Gambella, Fabrizio Pittorino, Manuel Roveri

Neural Architecture Search (NAS) paves the way for the automatic definition of Neural Network (NN) architectures, attracting increasing research attention and offering solutions in various scenarios. This study introduces a novel NAS solution, called Flat Neural Architecture Search (FlatNAS), which explores the interplay between a novel figure of merit based on robustness to weight perturbations and single NN optimization with Sharpness-Aware Minimization (SAM). FlatNAS is the first work in the literature to systematically explore flat regions in the loss landscape of NNs in a NAS procedure, while jointly optimizing their performance on in-distribution data, their out-of-distribution (OOD) robustness, and constraining the number of parameters in their architecture. Differently from current studies primarily concentrating on OOD algorithms, FlatNAS successfully evaluates the impact of NN architectures on OOD robustness, a crucial aspect in real-world applications of machine and deep learning. FlatNAS achieves a good trade-off between performance, OOD generalization, and the number of parameters, by using only in-distribution data in the NAS exploration. The OOD robustness of the NAS-designed models is evaluated by focusing on robustness to input data corruptions, using popular benchmark datasets in the literature.

Luca Colombo, Alessandro Falcetta, Manuel Roveri

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.

Francesco Puoti, Fabrizio Pittorino, Manuel Roveri

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.

Miguel Nogales, Matteo Gambella, Manuel Roveri

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.

Hazem Hesham Yousef Shalby, Manuel Roveri

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.

Fabrizio Pittorino, Manuel Roveri

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.

Hazem Hesham Yousef Shalby, Fabrizio Pittorino, Francesca Palermo et al.

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.

Mehmet Emre Akbulut, Hazem Hesham Yousef Shalby, Fabrizio Pittorino et al.

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).

Matteo Gambella, Manuel Roveri

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 maintaining a lower model size compared to prior methods. These results highlight the promise of combining NAS and selective expansion for efficient, adaptive learning in incremental scenarios.

Massimo Pavan, Claudio Galimberti, Manuel Roveri

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.

Hazem Hesham Yousef Shalby, Arianna De Vecchi, Alice Scandelli et al.

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.

Luca Colombo, Fabrizio Pittorino, Manuel Roveri

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.

Massimo Pavan, Gioele Mombelli, Francesco Sinacori et al.

TinyML is a novel area of machine learning that gained huge momentum in the last few years thanks to the ability to execute machine learning algorithms on tiny devices (such as Internet-of-Things or embedded systems). Interestingly, research in this area focused on the efficient execution of the inference phase of TinyML models on tiny devices, while very few solutions for on-device learning of TinyML models are available in the literature due to the relevant overhead introduced by the learning algorithms. The aim of this paper is to introduce a new type of adaptive TinyML solution that can be used in tasks, such as the presented \textit{Tiny Speaker Verification} (TinySV), that require to be tackled with an on-device learning algorithm. Achieving this goal required (i) reducing the memory and computational demand of TinyML learning algorithms, and (ii) designing a TinyML learning algorithm operating with few and possibly unlabelled training data. The proposed TinySV solution relies on a two-layer hierarchical TinyML solution comprising Keyword Spotting and Adaptive Speaker Verification module. We evaluated the effectiveness and efficiency of the proposed TinySV solution on a dataset collected expressly for the task and tested the proposed solution on a real-world IoT device (Infineon PSoC 62S2 Wi-Fi BT Pioneer Kit).

Lunchen Xie, Eugenio Lomurno, Matteo Gambella et al.

Accurate classification of medical images is essential for modern diagnostics. Deep learning advancements led clinicians to increasingly use sophisticated models to make faster and more accurate decisions, sometimes replacing human judgment. However, model development is costly and repetitive. Neural Architecture Search (NAS) provides solutions by automating the design of deep learning architectures. This paper presents ZO-DARTS+, a differentiable NAS algorithm that improves search efficiency through a novel method of generating sparse probabilities by bi-level optimization. Experiments on five public medical datasets show that ZO-DARTS+ matches the accuracy of state-of-the-art solutions while reducing search times by up to three times.

Matteo Gambella, Jary Pomponi, Simone Scardapane et al.

Early Exit Neural Networks (EENNs) endow astandard Deep Neural Network (DNN) with Early Exit Classifiers (EECs), to provide predictions at intermediate points of the processing when enough confidence in classification is achieved. This leads to many benefits in terms of effectiveness and efficiency. Currently, the design of EENNs is carried out manually by experts, a complex and time-consuming task that requires accounting for many aspects, including the correct placement, the thresholding, and the computational overhead of the EECs. For this reason, the research is exploring the use of Neural Architecture Search (NAS) to automatize the design of EENNs. Currently, few comprehensive NAS solutions for EENNs have been proposed in the literature, and a fully automated, joint design strategy taking into consideration both the backbone and the EECs remains an open problem. To this end, this work presents Neural Architecture Search for Hardware Constrained Early Exit Neural Networks (NACHOS), the first NAS framework for the design of optimal EENNs satisfying constraints on the accuracy and the number of Multiply and Accumulate (MAC) operations performed by the EENNs at inference time. In particular, this provides the joint design of backbone and EECs to select a set of admissible (i.e., respecting the constraints) Pareto Optimal Solutions in terms of best tradeoff between the accuracy and number of MACs. The results show that the models designed by NACHOS are competitive with the state-of-the-art EENNs. Additionally, this work investigates the effectiveness of two novel regularization terms designed for the optimization of the auxiliary classifiers of the EENN

Simone Disabato, Manuel Roveri

Tiny Machine Learning (TML) is a new research area whose goal is to design machine and deep learning techniques able to operate in Embedded Systems and IoT units, hence satisfying the severe technological constraints on memory, computation, and energy characterizing these pervasive devices. Interestingly, the related literature mainly focused on reducing the computational and memory demand of the inference phase of machine and deep learning models. At the same time, the training is typically assumed to be carried out in Cloud or edge computing systems (due to the larger memory and computational requirements). This assumption results in TML solutions that might become obsolete when the process generating the data is affected by concept drift (e.g., due to periodicity or seasonality effect, faults or malfunctioning affecting sensors or actuators, or changes in the users' behavior), a common situation in real-world application scenarios. For the first time in the literature, this paper introduces a Tiny Machine Learning for Concept Drift (TML-CD) solution based on deep learning feature extractors and a k-nearest neighbors classifier integrating a hybrid adaptation module able to deal with concept drift affecting the data-generating process. This adaptation module continuously updates (in a passive way) the knowledge base of TML-CD and, at the same time, employs a Change Detection Test to inspect for changes (in an active way) to quickly adapt to concept drift by removing the obsolete knowledge. Experimental results on both image and audio benchmarks show the effectiveness of the proposed solution, whilst the porting of TML-CD on three off-the-shelf micro-controller units shows the feasibility of what is proposed in real-world pervasive systems.

Patrizia Agnello, Silvia M. Ansaldi, Emilia Lenzi et al.

Extracting patterns and useful information from Natural Language datasets is a challenging task, especially when dealing with data written in a language different from English, like Italian. Machine and Deep Learning, together with Natural Language Processing (NLP) techniques have widely spread and improved lately, providing a plethora of useful methods to address both Supervised and Unsupervised problems on textual information. We propose RECKONition, a NLP-based system for Industrial Accidents at Work Prevention. RECKONition, which is meant to provide Natural Language Understanding, Clustering and Inference, is the result of a joint partnership with the Italian National Institute for Insurance against Accidents at Work (INAIL). The obtained results showed the ability to process textual data written in Italian describing industrial accidents dynamics and consequences.

Giuseppe Canonaco, Andrea Soprani, Manuel Roveri et al.

In most of the transfer learning approaches to reinforcement learning (RL) the distribution over the tasks is assumed to be stationary. Therefore, the target and source tasks are i.i.d. samples of the same distribution. In the context of this work, we consider the problem of transferring value functions through a variational method when the distribution that generates the tasks is time-variant, proposing a solution that leverages this temporal structure inherent in the task generating process. Furthermore, by means of a finite-sample analysis, the previously mentioned solution is theoretically compared to its time-invariant version. Finally, we will provide an experimental evaluation of the proposed technique with three distinct temporal dynamics in three different RL environments.

Simone Disabato, Alessandro Falcetta, Alessio Mongelluzzo et al.

Deep-learning-as-a-service is a novel and promising computing paradigm aiming at providing machine/deep learning solutions and mechanisms through Cloud-based computing infrastructures. Thanks to its ability to remotely execute and train deep learning models (that typically require high computational loads and memory occupation), such an approach guarantees high performance, scalability, and availability. Unfortunately, such an approach requires to send information to be processed (e.g., signals, images, positions, sounds, videos) to the Cloud, hence having potentially catastrophic-impacts on the privacy of users. This paper introduces a novel distributed architecture for deep-learning-as-a-service that is able to preserve the user sensitive data while providing Cloud-based machine and deep learning services. The proposed architecture, which relies on Homomorphic Encryption that is able to perform operations on encrypted data, has been tailored for Convolutional Neural Networks (CNNs) in the domain of image analysis and implemented through a client-server REST-based approach. Experimental results show the effectiveness of the proposed architecture.

Simone Disabato, Manuel Roveri, Cesare Alippi

Severe constraints on memory and computation characterizing the Internet-of-Things (IoT) units may prevent the execution of Deep Learning (DL)-based solutions, which typically demand large memory and high processing load. In order to support a real-time execution of the considered DL model at the IoT unit level, DL solutions must be designed having in mind constraints on memory and processing capability exposed by the chosen IoT technology. In this paper, we introduce a design methodology aiming at allocating the execution of Convolutional Neural Networks (CNNs) on a distributed IoT application. Such a methodology is formalized as an optimization problem where the latency between the data-gathering phase and the subsequent decision-making one is minimized, within the given constraints on memory and processing load at the units level. The methodology supports multiple sources of data as well as multiple CNNs in execution on the same IoT system allowing the design of CNN-based applications demanding autonomy, low decision-latency, and high Quality-of-Service.

Cesare Alippi, Giacomo Boracchi, Diego Carrera et al.

We address the problem of detecting changes in multivariate datastreams, and we investigate the intrinsic difficulty that change-detection methods have to face when the data dimension scales. In particular, we consider a general approach where changes are detected by comparing the distribution of the log-likelihood of the datastream over different time windows. Despite the fact that this approach constitutes the frame of several change-detection methods, its effectiveness when data dimension scales has never been investigated, which is indeed the goal of our paper. We show that the magnitude of the change can be naturally measured by the symmetric Kullback-Leibler divergence between the pre- and post-change distributions, and that the detectability of a change of a given magnitude worsens when the data dimension increases. This problem, which we refer to as \emph{detectability loss}, is due to the linear relationship between the variance of the log-likelihood and the data dimension. We analytically derive the detectability loss on Gaussian-distributed datastreams, and empirically demonstrate that this problem holds also on real-world datasets and that can be harmful even at low data-dimensions (say, 10).