Fabrizio Pittorino

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.

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.

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.

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.

Wilailuck Treerath, Fabrizio Pittorino

Active learning (AL) reduces labeling cost by querying informative samples, but in tabular settings its cold-start gains are often limited because uncertainty estimates are unreliable when models are trained on very few labels. Tabular foundation models such as TabPFN provide calibrated probabilistic predictions via in-context learning (ICL), i.e., without task-specific weight updates, enabling an AL regime in which the labeled context - rather than parameters - is iteratively optimized. We formalize Tabular Active In-Context Learning (Tab-AICL) and instantiate it with four acquisition rules: uncertainty (TabPFN-Margin), diversity (TabPFN-Coreset), an uncertainty-diversity hybrid (TabPFN-Hybrid), and a scalable two-stage method (TabPFN-Proxy-Hybrid) that shortlists candidates using a lightweight linear proxy before TabPFN-based selection. Across 20 classification benchmarks, Tab-AICL improves cold-start sample efficiency over retrained gradient-boosting baselines (CatBoost-Margin and XGBoost-Margin), measured by normalized AULC up to 100 labeled samples.

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.

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.

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).

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.

Brandon Livio Annesi, Clarissa Lauditi, Carlo Lucibello et al.

Empirical studies on the landscape of neural networks have shown that low-energy configurations are often found in complex connected structures, where zero-energy paths between pairs of distant solutions can be constructed. Here we consider the spherical negative perceptron, a prototypical non-convex neural network model framed as a continuous constraint satisfaction problem. We introduce a general analytical method for computing energy barriers in the simplex with vertex configurations sampled from the equilibrium. We find that in the over-parameterized regime the solution manifold displays simple connectivity properties. There exists a large geodesically convex component that is attractive for a wide range of optimization dynamics. Inside this region we identify a subset of atypical high-margin solutions that are geodesically connected with most other solutions, giving rise to a star-shaped geometry. We analytically characterize the organization of the connected space of solutions and show numerical evidence of a transition, at larger constraint densities, where the aforementioned simple geodesic connectivity breaks down.

Fabrizio Pittorino, Antonio Ferraro, Gabriele Perugini et al.

We systematize the approach to the investigation of deep neural network landscapes by basing it on the geometry of the space of implemented functions rather than the space of parameters. Grouping classifiers into equivalence classes, we develop a standardized parameterization in which all symmetries are removed, resulting in a toroidal topology. On this space, we explore the error landscape rather than the loss. This lets us derive a meaningful notion of the flatness of minimizers and of the geodesic paths connecting them. Using different optimization algorithms that sample minimizers with different flatness we study the mode connectivity and relative distances. Testing a variety of state-of-the-art architectures and benchmark datasets, we confirm the correlation between flatness and generalization performance; we further show that in function space flatter minima are closer to each other and that the barriers along the geodesics connecting them are small. We also find that minimizers found by variants of gradient descent can be connected by zero-error paths composed of two straight lines in parameter space, i.e. polygonal chains with a single bend. We observe similar qualitative results in neural networks with binary weights and activations, providing one of the first results concerning the connectivity in this setting. Our results hinge on symmetry removal, and are in remarkable agreement with the rich phenomenology described by some recent analytical studies performed on simple shallow models.

Carlo Lucibello, Fabrizio Pittorino, Gabriele Perugini et al.

Message-passing algorithms based on the Belief Propagation (BP) equations constitute a well-known distributed computational scheme. It is exact on tree-like graphical models and has also proven to be effective in many problems defined on graphs with loops (from inference to optimization, from signal processing to clustering). The BP-based scheme is fundamentally different from stochastic gradient descent (SGD), on which the current success of deep networks is based. In this paper, we present and adapt to mini-batch training on GPUs a family of BP-based message-passing algorithms with a reinforcement field that biases distributions towards locally entropic solutions. These algorithms are capable of training multi-layer neural networks with discrete weights and activations with performance comparable to SGD-inspired heuristics (BinaryNet) and are naturally well-adapted to continual learning. Furthermore, using these algorithms to estimate the marginals of the weights allows us to make approximate Bayesian predictions that have higher accuracy than point-wise solutions.

Fabrizio Pittorino, Carlo Lucibello, Christoph Feinauer et al.

The properties of flat minima in the empirical risk landscape of neural networks have been debated for some time. Increasing evidence suggests they possess better generalization capabilities with respect to sharp ones. First, we discuss Gaussian mixture classification models and show analytically that there exist Bayes optimal pointwise estimators which correspond to minimizers belonging to wide flat regions. These estimators can be found by applying maximum flatness algorithms either directly on the classifier (which is norm independent) or on the differentiable loss function used in learning. Next, we extend the analysis to the deep learning scenario by extensive numerical validations. Using two algorithms, Entropy-SGD and Replicated-SGD, that explicitly include in the optimization objective a non-local flatness measure known as local entropy, we consistently improve the generalization error for common architectures (e.g. ResNet, EfficientNet). An easy to compute flatness measure shows a clear correlation with test accuracy.

Carlo Baldassi, Fabrizio Pittorino, Riccardo Zecchina

Learning in Deep Neural Networks (DNN) takes place by minimizing a non-convex high-dimensional loss function, typically by a stochastic gradient descent (SGD) strategy. The learning process is observed to be able to find good minimizers without getting stuck in local critical points, and that such minimizers are often satisfactory at avoiding overfitting. How these two features can be kept under control in nonlinear devices composed of millions of tunable connections is a profound and far reaching open question. In this paper we study basic non-convex one- and two-layer neural network models which learn random patterns, and derive a number of basic geometrical and algorithmic features which suggest some answers. We first show that the error loss function presents few extremely wide flat minima (WFM) which coexist with narrower minima and critical points. We then show that the minimizers of the cross-entropy loss function overlap with the WFM of the error loss. We also show examples of learning devices for which WFM do not exist. From the algorithmic perspective we derive entropy driven greedy and message passing algorithms which focus their search on wide flat regions of minimizers. In the case of SGD and cross-entropy loss, we show that a slow reduction of the norm of the weights along the learning process also leads to WFM. We corroborate the results by a numerical study of the correlations between the volumes of the minimizers, their Hessian and their generalization performance on real data.