arXiv subjects

Claudio Angione

Publications and source records attributed to Claudio Angione.

11 recordsLinked to original sources

arXiv2025

Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments

The deployment of large-scale models, such as large language models (LLMs), incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalability and data security, there is a growing shift towards decentralized systems for model deployment, where choosing efficient inference acceleration schemes become crucial to manage computational resources effectively and enhance system responsiveness. In this work, we address the challenge of selecting optimal acceleration methods in decentralized systems by introducing a meta-learning-based framework. This framework automates the selection process by learning from historical performance data of various acceleration techniques across different tasks. Unlike traditional methods that rely on random selection or expert intuition, our approach systematically identifies the best acceleration strategies based on the specific characteristics of each task. We demonstrate that our meta-learning framework not only streamlines the decision-making process but also consistently outperforms conventional methods in terms of efficiency and performance. Our results highlight the potential of inference acceleration in decentralized AI systems, offering a path towards more democratic and economically feasible artificial intelligence solutions.

cs.LG↗

arXiv2025

Encrypted Large Model Inference: The Equivariant Encryption Paradigm

Large scale deep learning model, such as modern language models and diffusion architectures, have revolutionized applications ranging from natural language processing to computer vision. However, their deployment in distributed or decentralized environments raises significant privacy concerns, as sensitive data may be exposed during inference. Traditional techniques like secure multi-party computation, homomorphic encryption, and differential privacy offer partial remedies but often incur substantial computational overhead, latency penalties, or limited compatibility with non-linear network operations. In this work, we introduce Equivariant Encryption (EE), a novel paradigm designed to enable secure, "blind" inference on encrypted data with near zero performance overhead. Unlike fully homomorphic approaches that encrypt the entire computational graph, EE selectively obfuscates critical internal representations within neural network layers while preserving the exact functionality of both linear and a prescribed set of non-linear operations. This targeted encryption ensures that raw inputs, intermediate activations, and outputs remain confidential, even when processed on untrusted infrastructure. We detail the theoretical foundations of EE, compare its performance and integration complexity against conventional privacy preserving techniques, and demonstrate its applicability across a range of architectures, from convolutional networks to large language models. Furthermore, our work provides a comprehensive threat analysis, outlining potential attack vectors and baseline strategies, and benchmarks EE against standard inference pipelines in decentralized settings. The results confirm that EE maintains high fidelity and throughput, effectively bridging the gap between robust data confidentiality and the stringent efficiency requirements of modern, large scale model inference.

cs.CR↗

arXiv2024

Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments

The deployment of large-scale models, such as large language models (LLMs) and sophisticated image generation systems, incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalability and data security, there is a growing shift towards decentralized systems for deploying such models. In these decentralized environments, efficient inference acceleration becomes crucial to manage computational resources effectively and enhance system responsiveness. In this work, we address the challenge of selecting optimal acceleration methods in decentralized systems by introducing a meta-learning-based framework. This framework automates the selection process by learning from historical performance data of various acceleration techniques across different tasks. Unlike traditional methods that rely on random selection or expert intuition, our approach systematically identifies the best acceleration strategies based on the specific characteristics of each task. We demonstrate that our meta-learning framework not only streamlines the decision-making process but also consistently outperforms conventional methods in terms of efficiency and performance. Our results highlight the potential of meta-learning to revolutionize inference acceleration in decentralized AI systems, offering a path towards more democratic and economically feasible artificial intelligence solutions.

cs.LG↗

arXiv2024

Model Agnostic Hybrid Sharding For Heterogeneous Distributed Inference

The rapid growth of large-scale AI models, particularly large language models has brought significant challenges in data privacy, computational resources, and accessibility. Traditional centralized architectures often struggle to meet required data security and scalability needs which hinders the democratization of AI systems. Nesa introduces a model-agnostic sharding framework designed for decentralized AI inference. Our framework uses blockchain-based sequential deep neural network sharding to distribute computational tasks across a diverse network of nodes based on a personalised heuristic and routing mechanism. This enables efficient distributed training and inference for recent large-scale models even on consumer-grade hardware. We use compression techniques like dynamic blockwise quantization and mixed matrix decomposition to reduce data transfer and memory needs. We also integrate robust security measures, including hardware-based trusted execution environments to ensure data integrity and confidentiality. Evaluating our system across various natural language processing and vision tasks shows that these compression strategies do not compromise model accuracy. Our results highlight the potential to democratize access to cutting-edge AI technologies by enabling secure and efficient inference on a decentralized network.

cs.AI↗

arXiv2024

Towards Secure and Private AI: A Framework for Decentralized Inference

The rapid advancement of ML models in critical sectors such as healthcare, finance, and security has intensified the need for robust data security, model integrity, and reliable outputs. Large multimodal foundational models, while crucial for complex tasks, present challenges in scalability, reliability, and potential misuse. Decentralized systems offer a solution by distributing workload and mitigating central points of failure, but they introduce risks of unauthorized access to sensitive data across nodes. We address these challenges with a comprehensive framework designed for responsible AI development. Our approach incorporates: 1) Zero-knowledge proofs for secure model verification, enhancing trust without compromising privacy. 2) Consensus-based verification checks to ensure consistent outputs across nodes, mitigating hallucinations and maintaining model integrity. 3) Split Learning techniques that segment models across different nodes, preserving data privacy by preventing full data access at any point. 4) Hardware-based security through trusted execution environments (TEEs) to protect data and computations. This framework aims to enhance security and privacy and improve the reliability and fairness of multimodal AI systems. Promoting efficient resource utilization contributes to more sustainable AI development. Our state-of-the-art proofs and principles demonstrate the framework's effectiveness in responsibly democratizing artificial intelligence, offering a promising approach for building secure and private foundational models.

cs.CR↗

arXiv2022

A pipeline and comparative study of 12 machine learning models for text classification

Text-based communication is highly favoured as a communication method, especially in business environments. As a result, it is often abused by sending malicious messages, e.g., spam emails, to deceive users into relaying personal information, including online accounts credentials or banking details. For this reason, many machine learning methods for text classification have been proposed and incorporated into the services of most email providers. However, optimising text classification algorithms and finding the right tradeoff on their aggressiveness is still a major research problem. We present an updated survey of 12 machine learning text classifiers applied to a public spam corpus. A new pipeline is proposed to optimise hyperparameter selection and improve the models' performance by applying specific methods (based on natural language processing) in the preprocessing stage. Our study aims to provide a new methodology to investigate and optimise the effect of different feature sizes and hyperparameters in machine learning classifiers that are widely used in text classification problems. The classifiers are tested and evaluated on different metrics including F-score (accuracy), precision, recall, and run time. By analysing all these aspects, we show how the proposed pipeline can be used to achieve a good accuracy towards spam filtering on the Enron dataset, a widely used public email corpus. Statistical tests and explainability techniques are applied to provide a robust analysis of the proposed pipeline and interpret the classification outcomes of the 12 machine learning models, also identifying words that drive the classification results. Our analysis shows that it is possible to identify an effective machine learning model to classify the Enron dataset with an F-score of 94%.

cs.IR↗

arXiv2021

Using Machine Learning to Emulate Agent-Based Simulations

In this proof-of-concept work, we evaluate the performance of multiple machine-learning methods as statistical emulators for use in the analysis of agent-based models (ABMs). Analysing ABM outputs can be challenging, as the relationships between input parameters can be non-linear or even chaotic even in relatively simple models, and each model run can require significant CPU time. Statistical emulation, in which a statistical model of the ABM is constructed to facilitate detailed model analyses, has been proposed as an alternative to computationally costly Monte Carlo methods. Here we compare multiple machine-learning methods for ABM emulation in order to determine the approaches best suited to emulating the complex behaviour of ABMs. Our results suggest that, in most scenarios, artificial neural networks (ANNs) and gradient-boosted trees outperform Gaussian process emulators, currently the most commonly used method for the emulation of complex computational models. ANNs produced the most accurate model replications in scenarios with high numbers of model runs, although training times were longer than the other methods. We propose that agent-based modelling would benefit from using machine-learning methods for emulation, as this can facilitate more robust sensitivity analyses for the models while also reducing CPU time consumption when calibrating and analysing the simulation.

cs.MA↗

arXiv2020

Situating Agent-Based Modelling in Population Health Research

Today's most troublesome population health challenges are often driven by social and environmental determinants, which are difficult to model using traditional epidemiological methods. We agree with those who have argued for the wider adoption of agent-based modelling (ABM) in taking on these challenges. However, while ABM has been used occasionally in population health, we argue that for ABM to be most effective in the field it should be used as a means for answering questions normally inaccessible to the traditional epidemiological toolkit. In an effort to clearly illustrate the utility of ABM for population health research, and to clear up persistent misunderstandings regarding the method's conceptual underpinnings, we offer a detailed presentation of the core concepts of complex systems theory, and summarise why simulations are essential to the study of complex systems. We then examine the current state of the art in ABM for population health, and propose they are well-suited for the study of the `wicked' problems in population health, and could make significant contributions to theory and intervention development in these areas.

cs.MA↗

arXiv2018

Seeing the wood for the trees: a forest of methods for optimisation and omic-network integration in metabolic modelling

Metabolic modelling has entered a mature phase with dozens of methods and software implementations available to the practitioner and the theoretician. It is not easy for a modeller to be able to see the wood (or the forest) for the trees. Driven by this analogy, we here present a "forest" of principal methods used for constraint-based modelling in systems biology. This provides a tree-based view of methods available to prospective modellers, also available in interactive version at http://modellingmetabolism.net, where it will be kept updated with new methods after the publication of the present manuscript. Our updated classification of existing methods and tools highlights the most promising in the different branches, with the aim to develop a vision of how existing methods could hybridise and become more complex. We then provide the first hands-on tutorial for multi-objective optimisation of metabolic models in R. We finally discuss the implementation of multi-view machine-learning approaches in poly-omic integration. Throughout this work, we demonstrate the optimisation of trade-offs between multiple metabolic objectives, with a focus on omic data integration through machine learning. We anticipate that the combination of a survey, a perspective on multi-view machine learning, and a step-by-step R tutorial should be of interest for both the beginner and the advanced user.

q-bio.MN↗

arXiv2018

Integrating splice-isoform expression into genome-scale models characterizes breast cancer metabolism

Motivation: Despite being often perceived as the main contributors to cell fate and physiology, genes alone cannot predict cellular phenotype. During the process of gene expression, 95% of human genes can code for multiple proteins due to alternative splicing. While most splice variants of a gene carry the same function, variants within some key genes can have remarkably different roles. To bridge the gap between genotype and phenotype, condition- and tissue-specific models of metabolism have been constructed. However, current metabolic models only include information at the gene level. Consequently, as recently acknowledged by the scientific community, common situations where changes in splice-isoform expression levels alter the metabolic outcome cannot be modeled. Results: We here propose GEMsplice, the first method for the incorporation of splice-isoform expression data into genome-scale metabolic models. Using GEMsplice, we make full use of RNA-Seq quantitative expression profiles to predict, for the first time, the effects of splice isoform-level changes in the metabolism of 1455 patients with 31 different breast cancer types. We validate GEMsplice by generating cancer-versus-normal predictions on metabolic pathways, and by comparing with gene-level approaches and available literature on pathways affected by breast cancer. GEMsplice is freely available for academic use at https://github.com/GEMsplice/GEMsplice_code. Compared to state-of-the-art methods, we anticipate that GEMsplice will enable for the first time computational analyses at transcript level with splice-isoform resolution.

q-bio.GN↗

arXiv2013

Bose-Einstein Condensation in Satisfiability Problems

This paper is concerned with the complex behavior arising in satisfiability problems. We present a new statistical physics-based characterization of the satisfiability problem. Specifically, we design an algorithm that is able to produce graphs starting from a k-SAT instance, in order to analyze them and show whether a Bose-Einstein condensation occurs. We observe that, analogously to complex networks, the networks of k-SAT instances follow Bose statistics and can undergo Bose-Einstein condensation. In particular, k-SAT instances move from a fit-get-rich network to a winner-takes-all network as the ratio of clauses to variables decreases, and the phase transition of k-SAT approximates the critical temperature for the Bose-Einstein condensation. Finally, we employ the fitness-based classification to enhance SAT solvers (e.g., ChainSAT) and obtain the consistently highest performing SAT solver for CNF formulas, and therefore a new class of efficient hardware and software verification tools.

cs.DS↗