Domain generalization is the problem of assigning class labels to an unlabeled test data set, given several labeled training data sets drawn from similar distributions. This problem arises in several applications where data distributions fluctuate because of biological, technical, or other sources of variation. We develop a distribution-free, kernel-based approach that predicts a classifier from the marginal distribution of features, by leveraging the trends present in related classification tasks. This approach involves identifying an appropriate reproducing kernel Hilbert space and optimizing a regularized empirical risk over the space. We present generalization error analysis, describe universal kernels, and establish universal consistency of the proposed methodology. Experimental results on synthetic data and three real data applications demonstrate the superiority of the method with respect to a pooling strategy.

This paper presents novel Gaussian process decentralized data fusion algorithms exploiting the notion of agent-centric support sets for distributed cooperative perception of large-scale environmental phenomena. To overcome the limitations of scale in existing works, our proposed algorithms allow every mobile sensing agent to choose a different support set and dynamically switch to another during execution for encapsulating its own data into a local summary that, perhaps surprisingly, can still be assimilated with the other agents' local summaries (i.e., based on their current choices of support sets) into a globally consistent summary to be used for predicting the phenomenon. To achieve this, we propose a novel transfer learning mechanism for a team of agents capable of sharing and transferring information encapsulated in a summary based on a support set to that utilizing a different support set with some loss that can be theoretically bounded and analyzed. To alleviate the issue of information loss accumulating over multiple instances of transfer learning, we propose a new information sharing mechanism to be incorporated into our algorithms in order to achieve memory-efficient lazy transfer learning. Empirical evaluation on real-world datasets show that our algorithms outperform the state-of-the-art methods.

Transfer learning aims to faciliate learning tasks in a label-scarce target domain by leveraging knowledge from a related source domain with plenty of labeled data. Often times we may have multiple domains with little or no labeled data as targets waiting to be solved. Most existing efforts tackle target domains separately by modeling the `source-target' pairs without exploring the relatedness between them, which would cause loss of crucial information, thus failing to achieve optimal capability of knowledge transfer. In this paper, we propose a novel and effective approach called Multi-Relevance Transfer Learning (MRTL) for this purpose, which can simultaneously transfer different knowledge from the source and exploits the shared common latent factors between target domains. Specifically, we formulate the problem as an optimization task based on a collective nonnegative matrix tri-factorization framework. The proposed approach achieves both source-target transfer and target-target leveraging by sharing multiple decomposed latent subspaces. Further, an alternative minimization learning algorithm is developed with convergence guarantee. Empirical study validates the performance and effectiveness of MRTL compared to the state-of-the-art methods.

We consider the Hypothesis Transfer Learning (HTL) problem where one incorporates a hypothesis trained on the source domain into the learning procedure of the target domain. Existing theoretical analysis either only studies specific algorithms or only presents upper bounds on the generalization error but not on the excess risk. In this paper, we propose a unified algorithm-dependent framework for HTL through a novel notion of transformation function, which characterizes the relation between the source and the target domains. We conduct a general risk analysis of this framework and in particular, we show for the first time, if two domains are related, HTL enjoys faster convergence rates of excess risks for Kernel Smoothing and Kernel Ridge Regression than those of the classical non-transfer learning settings. Experiments on real world data demonstrate the effectiveness of our framework.

Despite increasing focus on data publication and discovery in materials science and related fields, the global view of materials data is highly sparse. This sparsity encourages training models on the union of multiple datasets, but simple unions can prove problematic as (ostensibly) equivalent properties may be measured or computed differently depending on the data source. These hidden contextual differences introduce irreducible errors into analyses, fundamentally limiting their accuracy. Transfer learning, where information from one dataset is used to inform a model on another, can be an effective tool for bridging sparse data while preserving the contextual differences in the underlying measurements. Here, we describe and compare three techniques for transfer learning: multi-task, difference, and explicit latent variable architectures. We show that difference architectures are most accurate in the multi-fidelity case of mixed DFT and experimental band gaps, while multi-task most improves classification performance of color with band gaps. For activation energies of steps in NO reduction, the explicit latent variable method is not only the most accurate, but also enjoys cancellation of errors in functions that depend on multiple tasks. These results motivate the publication of high quality materials datasets that encode transferable information, independent of industrial or academic interest in the particular labels, and encourage further development and application of transfer learning methods to materials informatics problems.

Transfer learning is a popular practice in deep neural networks, but fine-tuning of large number of parameters is a hard task due to the complex wiring of neurons between splitting layers and imbalance distributions of data in pretrained and transferred domains. The reconstruction of the original wiring for the target domain is a heavy burden due to the size of interconnections across neurons. We propose a distributed scheme that tunes the convolutional filters individually while backpropagates them jointly by means of basic probability assignment. Some of the most recent advances in evidence theory show that in a vast variety of the imbalanced regimes, optimizing of some proper objective functions derived from contingency matrices prevents biases towards high-prior class distributions. Therefore, the original filters get gradually transferred based on individual contributions to overall performance of the target domain. This largely reduces the expected complexity of transfer learning whilst highly improves precision. Our experiments on standard benchmarks and scenarios confirm the consistent improvement of our distributed deep transfer learning strategy.

An important goal of computer vision is to build systems that learn visual representations over time that can be applied to many tasks. In this paper, we investigate a vision-language embedding as a core representation and show that it leads to better cross-task transfer than standard multi-task learning. In particular, the task of visual recognition is aligned to the task of visual question answering by forcing each to use the same word-region embeddings. We show this leads to greater inductive transfer from recognition to VQA than standard multitask learning. Visual recognition also improves, especially for categories that have relatively few recognition training labels but appear often in the VQA setting. Thus, our paper takes a small step towards creating more general vision systems by showing the benefit of interpretable, flexible, and trainable core representations.

Today's leading minds talk AI with host Byron Reese Visit VoicesInAI.com to access the podcast, or subscribe now: Byron Reese: Hello, this is Voices in AI brought to you by Gigaom. I am your host, Byron Reese. Jeff is a Google Senior Fellow and he leads the Google Brain project. His work probably touches my life, and maybe yours, about every hour of every day, so I can't wait to begin the conversation. Welcome to the show Jeff. Jeff Dean: Hi Byron, this is Jeff Dean. I'm really good, Jeff, thanks for taking the time to chat. You went to work for Google, I believe, in the second millennium.

Highly configurable software systems, such as mobile apps, compilers, and big data engines, are increasingly exposed to end users and developers on a daily basis for varying use cases. Users are interested not only in the fastest configuration, but also in whether the fastest configuration for their applications also remains the fastest when the environmental situation has been changed. For instance, a mobile developer might be interested to know if the software that she has configured to consume minimal energy on a testing platform will also remain energy efficient on the users' mobile platform; or, in general, whether the configuration will remain optimal when the software is used in a different environment (e.g., with a different workload, on different hardware). Performance models have been extensively used to learn and describe the performance behavior of configurable systems [15], [19], [21], [23], [33], [43]-[45], [54], [61], [63]. However, the exponentially growing configuration space, complex interactions, and unknown constraints among configuration options [56] often make it costly and difficult to learn an accurate and reliable performance model. Even worse, existing techniques usually consider only a fixed environment (e.g., fixed workload, fixed hardware, fixed versions of the dependent libraries); should that environment change, a new performance model may need to be learned from scratch. This strong assumption limits the reusability of performance models across environments.

Tropical cyclone wind-intensity prediction is a challenging task considering drastic changes climate patterns over the last few decades. In order to develop robust prediction models, one needs to consider different characteristics of cyclones in terms of spatial and temporal characteristics. Transfer learning incorporates knowledge from a related source dataset to compliment a target datasets especially in cases where there is lack or data. Stacking is a form of ensemble learning focused for improving generalization that has been recently used for transfer learning problems which is referred to as transfer stacking. In this paper, we employ transfer stacking as a means of studying the effects of cyclones whereby we evaluate if cyclones in different geographic locations can be helpful for improving generalization performance. Moreover, we use conventional neural networks for evaluating the effects of duration on cyclones in prediction performance. Therefore, we develop an effective strategy that evaluates the relationships between different types of cyclones through transfer learning and conventional learning methods via neural networks.