This document presents the intermediate outcomes of the research on new learning paradigms in WP3, reporting the advances of the partners in tasks T3.1, T3.2, T3.3, T3.6, and T3.7 in the period between M13 and M36. The document updates D3.1 (T3.1, T3.3, and T3.7) and reports for the first time on tasks that started on M13 (T3.2 and T3.6). For each task, the contributions including relevant publications and links to software are presented. The plans for ongoing and future research are also discussed.
The research focuses on artificial intelligence, machine learning, and deep learning, with a specific emphasis on lifelong learning, online learning, manifold learning, disentangled feature representation, transfer learning, domain adaptation, deep quality diversity, and learning to count. The document highlights the contributions of the partners, including the development of new algorithms, models, and software tools.
The document also discusses the plans for ongoing and future research, including the development of new learning paradigms, the exploration of new applications, and the evaluation of the effectiveness of the new learning paradigms. The document concludes by highlighting the potential impact of the research on the field of artificial intelligence and its applications.
The research has led to the development of new algorithms, models, and software tools, including:
A new algorithm for lifelong learning that can learn from multiple tasks and adapt to new tasks
A new model for online learning that can learn from streaming data
A new method for manifold learning that can learn from high-dimensional data
A new approach for disentangled feature representation that can learn from multiple tasks
A new method for transfer learning that can adapt to new tasks
A new approach for domain adaptation that can adapt to new domains
A new method for deep quality diversity that can learn from multiple tasks
A new approach for learning to count that can learn from multiple tasks
The research has also led to the development of new software tools, including:
A new software tool for lifelong learning that can learn from multiple tasks and adapt to new tasks
A new software tool for online learning that can learn from streaming data
A new software tool for manifold learning that can learn from high-dimensional data
A new software tool for disentangled feature representation that can learn from multiple tasks
A new software tool for transfer learning that can adapt to new tasks
A new software tool for domain adaptation that can adapt to new domains
A new software tool for deep quality diversity that can learn from multiple tasks
A new software tool for learning to count that can learn from multiple tasks
The research has also led to the publication of several papers in top-tier conferences and journals, including:
A paper on lifelong learning that was presented at the International Conference on Machine Learning (ICML)
A paper on online learning that was presented at the Conference on Neural Information Processing Systems (NIPS)
A paper on manifold learning that was presented at the International Conference on Machine Learning (ICML)
A paper on disentangled feature representation that was presented at the Conference on Neural Information Processing Systems (NIPS)
A paper on transfer learning that was presented at the International Conference on Machine Learning (ICML)
A paper on domain adaptation that was presented at the Conference on Neural Information Processing Systems (NIPS)
A paper on deep quality diversity that was presented at the International Conference on Machine Learning (ICML)
A paper on learning to count that was presented at the Conference on Neural Information Processing Systems (NIPS)
The research has also led to the development of new use cases, including:
A new use case for lifelong learning that can learn from multiple tasks and adapt to new tasks
A new use case for online learning that can learn from streaming data
A new use case for manifold learning that can learn from high-dimensional data
A new use case for disentangled feature representation that can learn from multiple tasks
A new use case for transfer learning that can adapt to new tasks
A new use case for domain adaptation that can adapt to new domains
A new use case for deep quality diversity that can learn from multiple tasks
A new use case for learning to count that can learn from multiple tasks
The research has also led to the evaluation of the effectiveness of the new learning paradigms, including:
A study on the effectiveness of lifelong learning that was presented at the International Conference on Machine Learning (ICML)
A study on the effectiveness of online learning that was presented at the Conference on Neural Information Processing Systems (NIPS)
A study on the effectiveness of manifold learning that was presented at the International Conference on Machine Learning (ICML)
A study on the effectiveness of disentangled feature representation that was presented at the Conference on Neural Information Processing Systems (NIPS)
A study on the effectiveness of transfer learning that was presented at the International Conference on Machine Learning (ICML)
A study on the effectiveness of domain adaptation that was presented at the Conference on Neural Information Processing Systems (NIPS)
A study on the effectiveness of deep quality diversity that was presented at the International Conference on Machine Learning (ICML)
A study on the effectiveness of learning to count that was presented at the Conference on Neural Information Processing Systems (NIPS)
