Résultats de recherche pour « Quantum Machine Learning »

Identifying where quantum models may offer practical benefits in near term quantum machine learning (QML) requires moving beyond isolated algorithmic proposals toward systematic and empirical exploration across models, datasets, and hardware constraints. We introduce MerLin, an open-source framework designed as a discovery engine for photonic and hybrid quantum machine learning. MerLin integrates optimized strong simulation of linear optical circuits into standard PyTorch and scikit learn workflows, enabling end-to-end differentiable training of quantum layers. MerLin is designed around systematic benchmarking and reproducibility. As an initial contribution, we reproduce eighteen state-of-the-art photonic and hybrid QML works spanning kernel methods, reservoir computing, convolutional and recurrent architectures, generative models, and modern training paradigms. These reproductions are released as reusable, modular experiments that can be directly extended and adapted, establishing a shared experimental baseline consistent with empirical benchmarking methodologies widely adopted in modern artificial intelligence. By embedding photonic quantum models within established machine learning ecosystems, MerLin allows practitioners to leverage existing tooling for ablation studies, cross-modality comparisons, and hybrid classical-quantum workflows. The framework already implements hardware-aware features, allowing tests on available quantum hardware while enabling exploration beyond its current capabilities, positioning MerLin as a forward-looking co-design tool linking algorithms, benchmarks, and hardware.

Research at the intersection of machine learning (ML) and quantum physics is a recent growing field due to the enormous expectations and the success of both fields. ML is arguably one of the most promising technologies that has and will continue to disrupt many aspects of our lives. The way we do research is almost certainly no exception and ML, with its unprecedented ability to find hidden patterns in data, will be assisting future scientific discoveries. Quantum physics on the other side, even though it is sometimes not entirely intuitive, is one of the most successful physical theories and we are on the verge of adopting some quantum technologies in our daily life. Quantum many-body physics is a subfield of quantum physics where we study the collective behavior of particles or atoms and the emergence of phenomena that are due to this collective behavior, such as phases of matter. The study of phase transitions of these systems often requires some intuition of how we can quantify the order parameter of a phase. ML algorithms can imitate something similar to intuition by inferring knowledge from example data. They can, therefore, discover patterns that are invisible to the human eye, which makes them excellent candidates to study phase transitions. At the same time, quantum devices are known to be able to perform some computational task exponentially faster than classical computers and they are able to produce data patterns that are hard to simulate on classical computers. Therefore, there is the hope that ML algorithms run on quantum devices show an advantage over their classical analog. This thesis is devoted to study two different paths along the front lines of ML and quantum physics. On one side, we study the use of neural networks (NN) to classify phases of mater in many-body quantum systems. On the other side, we study ML algorithms that run on quantum computers. The connection between ML for quantum physics and quantum physics for ML in this thesis is an emerging subfield in ML, the interpretability of learning algorithms. A crucial ingredient in the study of phase transitions with NNs is a better understanding of the predictions of the NN, to eventually infer a model of the quantum system and interpretability can assist us in this endeavor. The interpretability method that we study analyzes the influence of the training points on a test prediction and it depends on the curvature of the NN loss landscape. This further inspired an in-depth study of the loss of quantum machine learning (QML) applications which we as well will discuss. In this thesis, we give answers to the questions of how we can leverage NNs to classify phases of matter and we use a method that allows to do domain adaptation to transfer the learned "intuition" from systems without noise onto systems with noise. To map the phase diagram of quantum many-body systems in a fully unsupervised manner, we study a method known from anomaly detection that allows us to reduce the human input to a mini mum. We will as well use interpretability methods to study NNs that are trained to distinguish phases of matter to understand if the NNs are learning something similar to an order parameter and if their way of learning can be made more accessible to humans. And finally, inspired by the interpretability of classical NNs, we develop tools to study the loss landscapes of variational quantum circuits to identify possible differences between classical and quantum ML algorithms that might be leveraged for a quantum advantage. La investigación en la intersección del aprendizaje automático (machine learning, ML) y la física cuántica es una área en crecimiento reciente debido al éxito y las enormes expectativas de ambas áreas. ML es posiblemente una de las tecnologías más prometedoras que ha alterado y seguirá alterando muchos aspectos de nuestras vidas. Es casi seguro que la forma en que investigamos no es una excepción y el ML, con su capacidad sin precedentes para encontrar patrones ocultos en los datos ayudará a futuros descubrimientos científicos. La física cuántica, por otro lado, aunque a veces no es del todo intuitiva, es una de las teorías físicas más exitosas, y además estamos a punto de adoptar algunas tecnologías cuánticas en nuestra vida diaria. La física cuántica de los muchos cuerpos (many-body) es una subárea de la física cuántica donde estudiamos el comportamiento colectivo de partículas o átomos y la aparición de fenómenos que se deben a este comportamiento colectivo, como las fases de la materia. El estudio de las transiciones de fase de estos sistemas a menudo requiere cierta intuición de cómo podemos cuantificar el parámetro de orden de una fase. Los algoritmos de ML pueden imitar algo similar a la intuición al inferir conocimientos a partir de datos de ejemplo. Por lo tanto, pueden descubrir patrones que son invisibles para el ojo humano, lo que los convierte en excelentes candidatos para estudiar las transiciones de fase. Al mismo tiempo, se sabe que los dispositivos cuánticos pueden realizar algunas tareas computacionales exponencialmente más rápido que los ordenadores clásicos y pueden producir patrones de datos que son difíciles de simular en los ordenadores clásicos. Por lo tanto, existe la esperanza de que los algoritmos ML que se ejecutan en dispositivos cuánticos muestren una ventaja sobre su analógico clásico. Estudiamos dos caminos diferentes a lo largo de la vanguardia del ML y la física cuántica. Por un lado, estudiamos el uso de redes neuronales (neural network, NN) para clasificar las fases de la materia en sistemas cuánticos de muchos cuerpos. Por otro lado, estudiamos los algoritmos ML que se ejecutan en ordenadores cuánticos. La conexión entre ML para la física cuántica y la física cuántica para ML en esta tesis es un subárea emergente en ML: la interpretabilidad de los algoritmos de aprendizaje. Un ingrediente crucial en el estudio de las transiciones de fase con NN es una mejor comprensión de las predicciones de la NN, para inferir un modelo del sistema cuántico. Así pues, la interpretabilidad de la NN puede ayudarnos en este esfuerzo. El estudio de la interpretabilitad inspiró además un estudio en profundidad de la pérdida de aplicaciones de aprendizaje automático cuántico (quantum machine learning, QML) que también discutiremos. En esta tesis damos respuesta a las preguntas de cómo podemos aprovechar las NN para clasificar las fases de la materia y utilizamos un método que permite hacer una adaptación de dominio para transferir la "intuición" aprendida de sistemas sin ruido a sistemas con ruido. Para mapear el diagrama de fase de los sistemas cuánticos de muchos cuerpos de una manera totalmente no supervisada, estudiamos un método conocido de detección de anomalías que nos permite reducir la entrada humana al mínimo. También usaremos métodos de interpretabilidad para estudiar las NN que están entrenadas para distinguir fases de la materia para comprender si las NN están aprendiendo algo similar a un parámetro de orden y si su forma de aprendizaje puede ser más accesible para los humanos. Y finalmente, inspirados por la interpretabilidad de las NN clásicas, desarrollamos herramientas para estudiar los paisajes de pérdida de los circuitos cuánticos variacionales para identificar posibles diferencias entre los algoritmos ML clásicos y cuánticos que podrían aprovecharse para obtener una ventaja cuántica.

Data science has become increasingly essential for the production of official statistics, as it enables the automated collection, processing, and analysis of large amounts of data. With such data science practices in place, it enables more timely, more insightful and more flexible reporting. However, the quality and integrity of data-science-driven statistics rely on the accuracy and reliability of the data sources and the machine learning techniques that support them. In particular, changes in data sources are inevitable to occur and pose significant risks that are crucial to address in the context of machine learning for official statistics. This paper gives an overview of the main risks, liabilities, and uncertainties associated with changing data sources in the context of machine learning for official statistics. We provide a checklist of the most prevalent origins and causes of changing data sources; not only on a technical level but also regarding ownership, ethics, regulation, and public perception. Next, we highlight the repercussions of changing data sources on statistical reporting. These include technical effects such as concept drift, bias, availability, validity, accuracy and completeness, but also the neutrality and potential discontinuation of the statistical offering. We offer a few important precautionary measures, such as enhancing robustness in both data sourcing and statistical techniques, and thorough monitoring. In doing so, machine learning-based official statistics can maintain integrity, reliability, consistency, and relevance in policy-making, decision-making, and public discourse.

Modern biology frequently relies on machine learning to provide predictions and improve decision processes. There have been recent calls for more scrutiny on machine learning performance and possible limitations. Here we present a set of community-wide recommendations aiming to help establish standards of supervised machine learning validation in biology. Adopting a structured methods description for machine learning based on data, optimization, model, evaluation (DOME) will aim to help both reviewers and readers to better understand and assess the performance and limitations of a method or outcome. The recommendations are formulated as questions to anyone wishing to pursue implementation of a machine learning algorithm. Answers to these questions can be easily included in the supplementary material of published papers.

Learning curves are a concept from social sciences that has been adopted in the context of machine learning to assess the performance of a learning algorithm with respect to a certain resource, e.g., the number of training examples or the number of training iterations. Learning curves have important applications in several machine learning contexts, most notably in data acquisition, early stopping of model training, and model selection. For instance, learning curves can be used to model the performance of the combination of an algorithm and its hyperparameter configuration, providing insights into their potential suitability at an early stage and often expediting the algorithm selection process. Various learning curve models have been proposed to use learning curves for decision making. Some of these models answer the binary decision question of whether a given algorithm at a certain budget will outperform a certain reference performance, whereas more complex models predict the entire learning curve of an algorithm. We contribute a framework that categorises learning curve approaches using three criteria: the decision-making situation they address, the intrinsic learning curve question they answer and the type of resources they use. We survey papers from the literature and classify them into this framework.

Online active learning is a paradigm in machine learning that aims to select the most informative data points to label from a data stream. The problem of minimizing the cost associated with collecting labeled observations has gained a lot of attention in recent years, particularly in real-world applications where data is only available in an unlabeled form. Annotating each observation can be time-consuming and costly, making it difficult to obtain large amounts of labeled data. To overcome this issue, many active learning strategies have been proposed in the last decades, aiming to select the most informative observations for labeling in order to improve the performance of machine learning models. These approaches can be broadly divided into two categories: static pool-based and stream-based active learning. Pool-based active learning involves selecting a subset of observations from a closed pool of unlabeled data, and it has been the focus of many surveys and literature reviews. However, the growing availability of data streams has led to an increase in the number of approaches that focus on online active learning, which involves continuously selecting and labeling observations as they arrive in a stream. This work aims to provide an overview of the most recently proposed approaches for selecting the most informative observations from data streams in real time. We review the various techniques that have been proposed and discuss their strengths and limitations, as well as the challenges and opportunities that exist in this area of research.