Reconciling Classical and Modern (Deep) Machine Learning for Real-World Applications

Descripción del proyecto

Modelos de aprendizaje automático para la modelización visual

El creciente número de parámetros resultante de un volumen de datos cada vez mayor está haciendo insostenible la tendencia actual de entrenar modelos predictivos de aprendizaje automático. En dicho contexto, el equipo del proyecto APHELEIA, financiado por el Consejo Europeo de Investigación, desarrollará modelos de aprendizaje automático más robustos, interpretables y eficientes, que exijan menos datos para obtener predicciones precisas, con especial atención a la modelización visual. El equipo del proyecto fomentará conexiones novedosas entre el procesamiento clásico de señales, la estadística, la optimización y el aprendizaje profundo moderno. Dicha sinergia permitirá a científicos e ingenieros concebir modelos de aprendizaje automático que puedan entrenarse e incorporen de forma inherente conocimientos previos sobre la semántica de las tareas y los procesos de generación de datos. Esos modelos darán prioridad automáticamente a las soluciones sencillas y coherentes frente a las intrincadas. El conjunto de herramientas informáticas de código abierto del proyecto puede aplicarse fácilmente a retos de reconocimiento visual e imagen inversa.

Objetivo

Despite the undeniable success of machine learning in addressing a wide variety of technological and scientific challenges, the current trend of training predictive models with an evergrowing number of parameters from an evergrowing amount of data is not sustainable. These huge models, often engineered by large corporations benefiting from huge computational resources, typically require learning a billion or more of parameters. They have proven to be very effective in solving prediction tasks in computer vision, natural language processing, and computational biology, for example, but they mostly remain black boxes that are hard to interpret, computationally demanding, and not robust to small data perturbations.
With a strong emphasis on visual modeling, the grand challenge of APHELEIA is to develop a new generation of machine learning models that are more robust, interpretable, and efficient, and do not require massive amounts of data to produce accurate predictions. To achieve this objective, we will foster new interactions between classical signal processing, statistics, optimization, and modern deep learning. Our goal is to reduce the need for massive data by enabling scientists and engineers to design trainable machine learning models that directly encode a priori knowledge of the task semantics and data formation process, while automatically prefering simple and stable solutions over complex ones. These models will be built on solid theoretical foundations with convergence and robustness guarantees, which are important to make real-life trustworthy predictions in the wild. We will implement these ideas in an open-source software toolbox readily applicable to visual recognition and inverse imaging problems, which will also handle other modalities. This will stimulate interdisciplinary collaborations, with the potential to be a game changer in the way scientists and engineers design machine learning problems.