Pietro Barbiero

• Master of Science in Mathematical Engineering
• Research Assistant at University of Cambridge

Concept Bottleneck Models (CBMs) are neural networks designed to conjoin high performance with ante-hoc interpretability. CBMs work by first mapping inputs (e.g., images) to high-level concepts (e.g., visible objects and their properties) and then use these to solve a downstream task (e.g., tagging or scoring an image) in an interpretable manner. T...

In this paper, we investigate how concept-based models (CMs) respond to out-of-distribution (OOD) inputs. CMs are interpretable neural architectures that first predict a set of high-level concepts (e.g., stripes, black) and then predict a task label from those concepts. In particular, we study the impact of concept interventions (i.e., operations w...

The most common methods in explainable artificial intelligence are post-hoc techniques which identify the most relevant features used by pretrained opaque models. Some of the most advanced post hoc methods can generate explanations that account for the mutual interactions of input features in the form of logic rules. However, these methods frequent...

Concept Bottleneck Models (CBMs) are machine learning models that improve interpretability by grounding their predictions on human-understandable concepts, allowing for targeted interventions in their decision-making process. However, when intervened on, CBMs assume the availability of humans that can identify the need to intervene and always provi...

Concept-based models are an emerging paradigm in deep learning that constrains the inference process to operate through human-interpretable concepts, facilitating explainability and human interaction. However, these architectures, on par with popular opaque neural models, fail to account for the true causal mechanisms underlying the target phenomen...

We formalize a novel modeling framework for achieving interpretability in deep learning, anchored in the principle of inference equivariance. While the direct verification of interpretability scales exponentially with the number of variables of the system, we show that this complexity can be mitigated by treating interpretability as a Markovian pro...

Human Sensing, a field that leverages technology to monitor human activities, psycho-physiological states, and interactions with the environment, enhances our understanding of human behavior and drives the development of advanced services that improve overall quality of life. However, its reliance on detailed and often privacy-sensitive data as the...

Clustering algorithms rely on complex optimisation processes that may be difficult to comprehend, especially for individuals who lack technical expertise. While many explainable artificial intelligence techniques exist for supervised machine learning, unsupervised learning -- and clustering in particular -- has been largely neglected. To complicate...

The lack of transparency in the decision-making processes of deep learning systems presents a significant challenge in modern artificial intelligence (AI), as it impairs users' ability to rely on and verify these systems. To address this challenge, Concept Bottleneck Models (CBMs) have made significant progress by incorporating human-interpretable...

Despite their success, Large-Language Models (LLMs) still face criticism as their lack of interpretability limits their controllability and reliability. Traditional post-hoc interpretation methods, based on attention and gradient-based analysis, offer limited insight into the model's decision-making processes. In the image field, Concept-based mode...

Causal opacity denotes the difficulty in understanding the "hidden" causal structure underlying a deep neural network's (DNN) reasoning. This leads to the inability to rely on and verify state-of-the-art DNN-based systems especially in high-stakes scenarios. For this reason, causal opacity represents a key open challenge at the intersection of deep...

Interpretable deep learning aims at developing neural architectures whose decision-making processes could be understood by their users. Among these techniqes, Concept Bottleneck Models enhance the interpretability of neural networks by integrating a layer of human-understandable concepts. These models, however, necessitate training a new model from...

Federated Learning (FL), a privacy-aware approach in distributed deep learning environments, enables many clients to collaboratively train a model without sharing sensitive data, thereby reducing privacy risks. However, enabling human trust and control over FL systems requires understanding the evolving behaviour of clients, whether beneficial or d...

Current deep learning models are not designed to simultaneously address three fundamental questions: predict class labels to solve a given classification task (the “What?”), explain task predictions (the “Why?”), and imagine alternative scenarios that could result in different predictions (the “What if?”). The inability to answer these questions re...

Deep learning has been recently used to extract the relevant features for representing input data also in the unsupervised setting. However, state-of-the-art techniques focus mostly on algorithmic efficiency and accuracy rather than mimicking the input manifold. On the contrary, competitive learning is a powerful tool for replicating the input dist...

Background Not being well controlled by therapy with inhaled corticosteroids and long‐acting β2 agonist bronchodilators is a major concern for severe‐asthma patients. The current treatment option for these patients is the use of biologicals such as anti‐IgE treatment, omalizumab, as an add‐on therapy. Despite the accepted use of omalizumab, patient...

The opaque reasoning of Graph Neural Networks induces a lack of human trust. Existing graph network explainers attempt to address this issue by providing post-hoc explanations, however, they fail to make the model itself more interpretable. To fill this gap, we introduce the Concept Distillation Module, the first differentiable concept-distillation...

The design of interpretable deep learning models working in relational domains poses an open challenge: interpretable deep learning methods, such as Concept-Based Models (CBMs), are not designed to solve relational problems, while relational models are not as interpretable as CBMs. To address this problem, we propose Relational Concept-Based Models...

Deep clustering is a branch of deep learning, in which the dimensionality reduction capabilities of deep networks are exploited for clustering data. In this sense, the deep part of clustering has to be considered more as a preprocessing step than a perfectly integrated module of the neural network. This paper proposes the idea of dual neural networ...

The opaqueness of deep neural networks hinders their employment in safety-critical applications. This issue is driving the research community to focus on eXplainable Artificial Intelligence (XAI) techniques. XAI algorithms can be categorized into two types: those that explain the predictions of black-box models and those that create interpretable m...

Multimodal learning is an essential paradigm for addressing complex real-world problems, where individual data modalities are typically insufficient to accurately solve a given modelling task. While various deep learning approaches have successfully addressed these challenges, their reasoning process is often opaque; limiting the capabilities for a...

Graph neural networks (GNNs) are highly effective on a variety of graph-related tasks; however, they lack interpretability and transparency. Current explainability approaches are typically local and treat GNNs as black-boxes. They do not look inside the model, inhibiting human trust in the model and explanations. Motivated by the ability of neurons...

Recent work on interpretability has focused on concept-based explanations, where deep learning models are explained in terms of high-level units of information, referred to as concepts. Concept learning models, however, have been shown to be prone to encoding impurities in their representations, failing to fully capture meaningful features of their...

The rise of Artificial Intelligence (AI) recently empowered researchers to investigate hard mathematical problems which eluded traditional approaches for decades. Yet, the use of AI in Universal Algebra (UA)--one of the fields laying the foundations of modern mathematics--is still completely unexplored. This work proposes the first use of AI to inv...

Deep learning methods are highly accurate, yet their opaque decision process prevents them from earning full human trust. Concept-based models aim to address this issue by learning tasks based on a set of human-understandable concepts. However, state-of-the-art concept-based models rely on high-dimensional concept embedding representations which la...

Explainable AI (XAI) aims to answer ethical and legal questions associated with the deployment of AI models. However, a considerable number of domain-specific reviews highlight the need of a mathematical foundation for the key notions in the field, considering that even the term "explanation" still lacks a precise definition. These reviews also adv...

Explainable AI (XAI) underwent a recent surge in research on concept extraction, focusing on extracting human-interpretable concepts from Deep Neural Networks. An important challenge facing concept extraction approaches is the difficulty of interpreting and evaluating discovered concepts, especially for complex tasks such as molecular property pred...

Recent work on interpretability has focused on concept-based explanations, where deep learning models are explained in terms of high-level units of information, referred to as concepts. Concept learning models, however, have been shown to be prone to encoding impurities in their representations, failing to fully capture meaningful features of their...

The large and still increasing popularity of deep learning clashes with a major limit of neural network architectures, that consists in their lack of capability in providing human-understandable motivations of their decisions. In situations in which the machine is expected to support the decision of human experts, providing a comprehensible explana...

Recently, Logic Explained Networks (LENs) have been proposed as explainable-by-design neural models providing logic explanations for their predictions. However, these models have only been applied to vision and tabular data, and they mostly favour the generation of global explanations, while local ones tend to be noisy and verbose. For these reason...

While instance-level explanation of GNN is a well-studied problem with plenty of approaches being developed, providing a global explanation for the behaviour of a GNN is much less explored, despite its potential in interpretability and debugging. Existing solutions either simply list local explanations for a given class, or generate a synthetic pro...

Deploying AI-powered systems requires trustworthy models supporting effective human interactions, going beyond raw prediction accuracy. Concept bottleneck models promote trustworthiness by conditioning classification tasks on an intermediate level of human-like concepts. This enables human interventions which can correct mispredicted concepts to im...

Graph neural networks (GNNs) are highly effective on a variety of graph-related tasks; however, they lack interpretability and transparency. Current explainability approaches are typically local and treat GNNs as black-boxes. They do not look inside the model, inhibiting human trust in the model and explanations. Motivated by the ability of neurons...

The opaque reasoning of Graph Neural Networks induces a lack of human trust. Existing graph network explainers attempt to address this issue by providing post-hoc explanations, however, they fail to make the model itself more interpretable. To fill this gap, we introduce the Concept Encoder Module, the first differentiable concept-discovery approac...

Explainable artificial intelligence has rapidly emerged since lawmakers have started requiring interpretable models for safety-critical domains. Concept-based neural networks have arisen as explainable-by-design methods as they leverage human-understandable symbols (i.e. concepts) to predict class memberships. However, most of these approaches focu...

Recent research on graph neural network (GNN) models successfully applied GNNs to classical graph algorithms and combinatorial optimisation problems. This has numerous benefits, such as allowing applications of algorithms when preconditions are not satisfied, or reusing learned models when sufficient training data is not available or can't be gener...

We propose an unsupervised, model-agnostic, wrapper method for feature selection. We assume that if a feature can be predicted using the others, it adds little information to the problem, and therefore could be removed without impairing the performance of whatever model will be eventually built. The proposed method iteratively identifies and remove...

Due to the current high availability of omics, data-driven biology has greatly expanded, and several papers have reviewed state-of-the-art technologies. Nowadays, two main types of investigation are available for a multi-omics dataset: extraction of relevant features for a meaningful biological interpretation and clustering of the samples. In the l...

Objective: Modern medicine needs to shift from a wait and react, curative discipline to a preventative, interdisciplinary science aiming at providing personalized, systemic, and precise treatment plans to patients. To this purpose, we propose a “digital twin” of patients modeling the human body as a whole and providing a panoramic view over individ...

The large and still increasing popularity of deep learning clashes with a major limit of neural network architectures, that consists in their lack of capability in providing human-understandable motivations of their decisions. In situations in which the machine is expected to support the decision of human experts, providing a comprehensible explana...

Topological learning is a wide research area aiming at uncovering the mutual spatial relationships between the elements of a set. Some of the most common and oldest approaches involve the use of unsupervised competitive neural networks. However, these methods are not based on gradient optimization which has been proven to provide striking results i...

Recent research on graph neural network (GNN) models successfully applied GNNs to classical graph algorithms and combinatorial optimisation problems. This has numerous benefits, such as allowing applications of algorithms when preconditions are not satisfied, or reusing learned models when sufficient training data is not available or can't be gener...

Explainable artificial intelligence has rapidly emerged since lawmakers have started requiring interpretable models for safety-critical domains. Concept-based neural networks have arisen as explainable-by-design methods as they leverage human-understandable symbols (i.e. concepts) to predict class memberships. However, most of these approaches focu...

LENs is a Python module integrating a variety of state-of-the-art approaches to provide logic explanations from neural networks. This package focuses on bringing these methods to non-specialists. It has minimal dependencies and it is distributed under the Apache 2.0 licence allowing both academic and commercial use. Source code and documentation ca...

Facial Emotion Recognition (FER) is the automatic processing of human emotions by means of facial expression analysis [1]. The most common approach exploits 3D Face Descriptors (3D-FD) [2], which derive from depth maps [3] by using mathematical operators. In recent years, Convolutional Neural Networks (CNNs) have been successfully employed in a wid...

In the field of machine learning, coresets are defined as subsets of the training set that can be used to obtain a good approximation of the behavior that a given algorithm would have on the whole training set. Advantages of using coresets instead of the training set include improving training speed and allowing for a better human understanding of...

Despite all the progress made in biomedical field, the Electrocardiogram (ECG) is still one of the most commonly used signal used in medical examinations. The problem of ECG classification has been approached in many different ways. Most of them rely on the extraction of features from the signal in the form of temporal or morphological characterist...

In recent years, due to the high availability of omic data, data driven biology has greatly expanded. However, the analysis of different data sources is still an open challenge. A few multi-omic approaches have been proposed in literature. However, none of them take into consideration the intrinsic topology of each omic. In this work, an unsupervis...

Objective: Modern medicine needs to shift from a wait and react, curative discipline to a preventative, interdisciplinary science aiming at providing personalised, systemic and precise treatment plans to patients. The aim of this work is to present how the integration of machine learning approaches with mechanistic computational modelling could yie...

Deep learning has been widely used for supervised learning and classification/regression problems. Recently, a novel area of research has applied this paradigm to unsupervised tasks; indeed, a gradient-based approach extracts, efficiently and autonomously, the relevant features for handling input data. However, state-of-the-art techniques focus mos...

Topological learning is a wide research area aiming at uncovering the mutual spatial relationships between the elements of a set. Some of the most common and oldest approaches involve the use of unsupervised competitive neural networks. However, these methods are not based on gradient optimization which has been proven to provide striking results i...

As machine learning becomes more and more available to the general public, theoretical questions are turning into pressing practical issues. Possibly, one of the most relevant concerns is the assessment of our confidence in trusting machine learning predictions. In many real-world cases, it is of utmost importance to estimate the capabilities of a...

A bstract Medicine is moving from reacting to a disease to prepare personalised and precision paths to well being. The complex and multi level pathophysiological patterns of most diseases require a systemic medicine approach and are challenging current medical therapies. Computational medicine is a vibrant interdisciplinary field that could help mo...

Feature selection is the process of choosing, or removing, features to obtain the most informative feature subset of minimal size. Such subsets are used to improve performance of machine learning algorithms and enable human understanding of the results. Approaches to feature selection in literature exploit several optimization algorithms. Multi-obj...

A coreset is a subset of the training set, using which a machine learning algorithm obtains performances similar to what it would deliver if trained over the whole original data. Coreset discovery is an active and open line of research as it allows improving training speed for the algorithms and may help human understanding the results. Building on...

Machine learning agents learn to take decisions extracting information from training data. When similar inferences can be obtained using a small subset of the same training set of samples, the subset is called coreset. Coresets discovery is an active line of research as it may be used to reduce the training speed as well as to allow human experts t...

In the field of artificial intelligence, agents learn how to take decisions by fitting their parameters on a set of samples called training set. Similarly, a core set is a subset of the training samples such that, if an agent exploits this set to fit its parameters instead of the whole training set, then the quality of the inferences does not chang...

In pattern recognition, neural networks can be used not only for the classification task, but also for feature selection and other intermediate steps. This paper addresses the 3D face recognition problem in order to select the most meaningful geometric descriptors. At this aim, the classification results are directly integrated in a biclustering pr...

When a machine learning algorithm is able to obtain the same performance given a complete training set, and a small subset of samples from the same training set, the subset is termed coreset. As using a coreset improves training speed and allows human experts to gain a better understanding of the data, by reducing the number of samples to be examin...

In machine learning a coreset is defined as a subset of the training set using which an algorithm obtains performances similar to what it would deliver if trained over the whole original data. Advantages of coresets include improving training speed and easing human understanding. Coreset discovery is an open line of research as limiting the trainin...

Hierarchical clustering is an important tool for extracting information from data in a multi-resolution way. It is more meaningful if driven by data, as in the case of divisive algorithms, which split data until no more division is allowed. However, they have the drawback of the splitting threshold setting. The neural networks can address this prob...

In an optimization problem, a coreset can be defined as a subset of the input points, such that a good approximation to the optimization problem can be obtained by solving it directly on the coreset, instead of using the whole original input. In machine learning, coresets are exploited for applications ranging from speeding up training time, to hel...

In an optimization problem, a coreset can be defined as a subset of the input points, such that a good approximation to the optimization problem can be obtained by solving it directly on the coreset, instead of using the whole original input. In machine learning, coresets are exploited for applications ranging from speeding up training time, to hel...

Cancer is a large family of genetic diseases that involve abnormal cell growth. Genetic mutations can vary from one patient to another. Therefore personalized precision is required to increase the reliability of prognostic predictions and the benefit of therapeutic decisions. The most important issues concerning gene analysis are strong noise, high...

Cancer is a large family of genetic diseases that involve abnormal cell growth. Genetic mutations can vary from one patient to another. Therefore personalized precision is required to increase the reliability of prognostic predictions and the benefit of therapeutic decisions. The most important issues concerning gene analysis are strong noise, high...

The analysis of complex systems, such as cancer resistance to drugs, require flexible algorithms but also simple models, as they will be used by biologists in order to get insights on the underlying phenomenon. Exploiting the availability of the largest collection of Patient-Derived Xenografts from metastatic colorectal cancer annotated for respons...

Exploiting the availability of the largest collection of Patient-Derived Xenografts from metastatic colorectal cancer annotated for response to therapies, this manuscript aims to characterize the biological phenomenon from a mathematical point of view. In particular, we design an experiment in order to investigate how genes interact with each other...

Presentation of the conference paper "Neural Epistemology in Time Series Learning". The video of the presentation is available at this link: https://youtu.be/PgC3a8w1obY.

In the last few years neural networks are effectively applied in different fields. However, the application of empirical-like algorithms as feed-forward neural networks is not always justified from an epistemological point of view. In this work the assumptions for the appropriate application of machine learning empirical-like algorithms to dynamica...

In pattern recognition, neural networks can be used not only for the classification task, but also for feature selection and other intermediate steps. This paper addresses the 3D face recognition problem in order to select the most meaningful geometric descriptors. At this aim, the classification results are directly integrated in a biclustering pr...

Presentation of the conference papers: - "DNA microarray classification: a Shallow Neural Network Model" - "‘DNA microarray classification: Evolutionary Optimization of Neural Network Hyperparameters"

This study investigates the relationship between the level to which a person feels connected to Nature and that person’s ability to perceive the restorative value of a natural environment. We assume that perceived restorativeness may depend on an individual’s connection to Nature and this relationship may also vary with the biophilic quality of the...

Clustering in high dimensional spaces is a very difficult task. Dealing with DNA microarrays is even more difficult because gene subsets are coregulated and coexpressed only under specific conditions. Biclusterng addresses the problem of finding such submanifolds by exploiting both gene and condition (tissue) clustering. The paper proposes a selfor...

Presentation of the conference papers: - "Supervised Gene Identification in Colorectal Cancer" - "‘Unsupervised Gene Identification in Colorectal Cancer"