It has been recently observed that a transfer learning solution might be all we needed to solve many few-shot learning benchmarks. This raises important questions about when and how meta-learning algorithms should be deployed. In this paper, we make a first step in clarifying these questions by first formulating a computable metric for a few-shot learning benchmark that we hypothesize is predictive of whether meta-learning solutions will succeed or not. We name this metric the diversity coefficient of a few-shot learning benchmark. Using the diversity coefficient, we show that the MiniImagenet benchmark has zero diversity - according to twenty-four different ways to compute the diversity. We proceed to show that when making a fair comparison between MAML learned solutions to transfer learning, both have identical meta-test accuracy. This suggests that transfer learning fails to outperform MAML - contrary to what previous work suggests. Together, these two facts provide the first test of whether diversity correlates with meta-learning success and therefore show that a diversity coefficient of zero correlates with a high similarity between transfer learning and MAML learned solutions - especially at meta-test time. We therefore conjecture meta-learned solutions have the same meta-test performance as transfer learning when the diversity coefficient is zero.

A key assumption in supervised learning is that training and test data follow the same probability distribution. However, this fundamental assumption is not always satisfied in practice, e.g., due to changing environments, sample selection bias, privacy concerns, or high labeling costs. Transfer learning (TL) relaxes this assumption and allows us to learn under distribution shift. Classical TL methods typically rely on importance-weighting -- a predictor is trained based on the training losses weighted according to the importance (i.e., the test-over-training density ratio). However, as real-world machine learning tasks are becoming increasingly complex, high-dimensional, and dynamical, novel approaches are explored to cope with such challenges recently. In this article, after introducing the foundation of TL based on importance-weighting, we review recent advances based on joint and dynamic importance-predictor estimation. Furthermore, we introduce a method of causal mechanism transfer that incorporates causal structure in TL. Finally, we discuss future perspectives of TL research.

Existing research on continual learning of a sequence of tasks focused on dealing with catastrophic forgetting, where the tasks are assumed to be dissimilar and have little shared knowledge. Some work has also been done to transfer previously learned knowledge to the new task when the tasks are similar and have shared knowledge. To the best of our knowledge, no technique has been proposed to learn a sequence of mixed similar and dissimilar tasks that can deal with forgetting and also transfer knowledge forward and backward. This paper proposes such a technique to learn both types of tasks in the same network. For dissimilar tasks, the algorithm focuses on dealing with forgetting, and for similar tasks, the algorithm focuses on selectively transferring the knowledge learned from some similar previous tasks to improve the new task learning. Additionally, the algorithm automatically detects whether a new task is similar to any previous tasks. Empirical evaluation using sequences of mixed tasks demonstrates the effectiveness of the proposed model.

We provide a regularization framework for subject transfer learning in which we seek to train an encoder and classifier to minimize classification loss, subject to a penalty measuring independence between the latent representation and the subject label. We introduce three notions of independence and corresponding penalty terms using mutual information or divergence as a proxy for independence. For each penalty term, we provide several concrete estimation algorithms, using analytic methods as well as neural critic functions. We provide a hands-off strategy for applying this diverse family of regularization algorithms to a new dataset, which we call "AutoTransfer". We evaluate the performance of these individual regularization strategies and our AutoTransfer method on EEG, EMG, and ECoG datasets, showing that these approaches can improve subject transfer learning for challenging real-world datasets. In this work, we investigate methods for transfer learning in the classification of biosignals data. Previous work has established the difficulty of transfer learning for biosignals and even the issue of so-called "negative transfer", in which naive attempts to combine datasets from multiple subjects or sessions can paradoxically decrease model performance, due to differences in response statistics [1, 2]. We address the problem of subject transfer by training models to be invariant to changes in a nuisance variable representing subject identifier.

With growing societal acceptance and increasing cost efficiency due to mass production, service robots are beginning to cross from the industrial to the social domain. Currently, customer service robots tend to be digital and emulate social interactions through on-screen text, but state-of-the-art research points towards physical robots soon providing customer service in person. This article explores two possibilities. Firstly, whether transfer learning can aid in the improvement of customer service chatbots between business domains. Secondly, the implementation of a framework for physical robots for in-person interaction. Modelled on social interaction with customer support Twitter accounts, transformer-based chatbot models are initially tasked to learn one domain from an initial random weight distribution. Given shared vocabulary, each model is then tasked with learning another domain by transferring knowledge from the prior. Following studies on 19 different businesses, results show that the majority of models are improved when transferring weights from at least one other domain, in particular those that are more data-scarce than others. General language transfer learning occurs, as well as higher-level transfer of similar domain knowledge in several cases. The chatbots are finally implemented on Temi and Pepper robots, with feasibility issues encountered and solutions are proposed to overcome them.

Object detection is a central downstream task used to test if pre-trained network parameters confer benefits, such as improved accuracy or training speed. The complexity of object detection methods can make this benchmarking non-trivial when new architectures, such as Vision Transformer (ViT) models, arrive. These difficulties (e.g., architectural incompatibility, slow training, high memory consumption, unknown training formulae, etc.) have prevented recent studies from benchmarking detection transfer learning with standard ViT models. In this paper, we present training techniques that overcome these challenges, enabling the use of standard ViT models as the backbone of Mask R-CNN. These tools facilitate the primary goal of our study: we compare five ViT initializations, including recent state-of-the-art self-supervised learning methods, supervised initialization, and a strong random initialization baseline. Our results show that recent masking-based unsupervised learning methods may, for the first time, provide convincing transfer learning improvements on COCO, increasing box AP up to 4% (absolute) over supervised and prior self-supervised pre-training methods. Moreover, these masking-based initializations scale better, with the improvement growing as model size increases.

Motivation: In recent years, image-based biological assays have steadily become high-throughput, sparking a need for fast automated methods to extract biologically-meaningful information from hundreds of thousands of images. Taking inspiration from the success of ImageNet, we curate CytoImageNet, a large-scale dataset of openly-sourced and weakly-labeled microscopy images (890K images, 894 classes). Pretraining on CytoImageNet yields features that are competitive to ImageNet features on downstream microscopy classification tasks. We show evidence that CytoImageNet features capture information not available in ImageNet-trained features. The dataset is made available at https://www.kaggle.com/stanleyhua/cytoimagenet.

Bayesian optimization is a powerful paradigm to optimize black-box functions based on scarce and noisy data. Its data efficiency can be further improved by transfer learning from related tasks. While recent transfer models meta-learn a prior based on large amount of data, in the low-data regime methods that exploit the closed-form posterior of Gaussian processes (GPs) have an advantage. In this setting, several analytically tractable transfer-model posteriors have been proposed, but the relative advantages of these methods are not well understood. In this paper, we provide a unified view on hierarchical GP models for transfer learning, which allows us to analyze the relationship between methods. As part of the analysis, we develop a novel closed-form boosted GP transfer model that fits between existing approaches in terms of complexity. We evaluate the performance of the different approaches in large-scale experiments and highlight strengths and weaknesses of the different transfer-learning methods.

Neural Painters is a class of models that follows a GAN framework to generate brushstrokes, which are then composed to create paintings. GANs are great generative models for AI Art but they are known to be notoriously difficult to train. To overcome GAN's limitations and to speed up the Neural Painter training, we applied Transfer Learning to the process reducing it from days to only hours, while achieving the same level of visual aesthetics in the final paintings generated. We report our approach and results in this work.

The recent abundance of data on electricity consumption at different scales opens new challenges and highlights the need for new techniques to leverage information present at finer scales in order to improve forecasts at wider scales. In this work, we take advantage of the similarity between this hierarchical prediction problem and multi-scale transfer learning. We develop two methods for hierarchical transfer learning, based respectively on the stacking of generalized additive models and random forests, and on the use of aggregation of experts. We apply these methods to two problems of electricity load forecasting at national scale, using smart meter data in the first case, and regional data in the second case. For these two usecases, we compare the performances of our methods to that of benchmark algorithms, and we investigate their behaviour using variable importance analysis. Our results demonstrate the interest of both methods, which lead to a significant improvement of the predictions.