Domain adaptation aims to mitigate distribution shifts among different domains. However, traditional formulations are mostly limited to categorical domains, greatly simplifying nuanced domain relationships in the real world. In this work, we tackle a generalization with taxonomy-structured domains, which formalizes domains with nested, hierarchical similarity structures such as animal species and product catalogs. We build on the classic adversarial framework and introduce a novel taxonomist, which competes with the adversarial discriminator to preserve the taxonomy information. The equilibrium recovers the classic adversarial domain adaptation’s solution if given a non-informative domain taxonomy (e.g., a flat taxonomy where all leaf nodes connect to the root node) while yielding non-trivial results with other taxonomies. Empirically, our method achieves state-of-the-art performance on both synthetic and real-world datasets with successful adaptation.

Understanding users' behavior at home is central to behavioral research. For example, social researchers are interested in studying domestic abuse, and healthcare professionals are interested in caregiver-patient interaction. Today, such studies rely on diaries and questionnaires, which are subjective, erroneous, and hard to sustain in longitudinal studies. We introduce Marko, a system that automatically collects behavior-related data, without asking people to write diaries or wear sensors. Marko transmits a low power wireless signal and analyses its reflections from the environment. It maps those reflections to how users interact with the environment (e.g., access to medication cabinet) and with each other (e.g., watch TV together). It provides novel algorithms for identifying who-does-what, and bootstrapping the system in new homes without asking users for new annotations. We evaluate Marko with a one-month deployment in six homes, and demonstrate its value for studying couple relationships and caregiver-patient interaction.

We provide a new approach to training neural models to exhibit transparency in a well-defined, functional manner. Our approach naturally operates over structured data and tailors the predictor, functionally, towards a chosen family of (local) witnesses. The estimation problem is setup as a co-operative game between an unrestricted predictor such as a neural network, and a set of witnesses chosen from the desired transparent family. The goal of the witnesses is to highlight, locally, how well the predictor conforms to the chosen family of functions, while the predictor is trained to minimize the highlighted discrepancy. We emphasize that the predictor remains globally powerful as it is only encouraged to agree locally with locally adapted witnesses. We analyze the effect of the proposed approach, provide example formulations in the context of deep graph and sequence models, and empirically illustrate the idea in chemical property prediction, temporal modeling, and molecule representation learning.

Strong theoretical guarantees of robustness can be given for ensembles of classifiers generated by input randomization. Specifically, an $\ell_2$ bounded adversary cannot alter the ensemble prediction generated by an additive isotropic Gaussian noise, where the radius for the adversary depends on both the variance of the distribution as well as the ensemble margin at the point of interest. We build on and considerably expand this work across broad classes of distributions. In particular, we offer adversarial robustness guarantees and associated algorithms for the discrete case where the adversary is $\ell_0$ bounded. Moreover, we exemplify how the guarantees can be tightened with specific assumptions about the function class of the classifier such as a decision tree. We empirically illustrate these results with and without functional restrictions across image and molecule datasets.

This paper proposes to address the word sense ambiguity issue in an unsupervised manner, where word sense representations are learned along a word sense selection mechanism given contexts.Prior work focused on designing a single model to deliver both mechanisms, and thus suffered from either coarse-grained representation learning or inefficient sense selection.The proposed modular approach, MUSE, implements flexible modules to optimize distinct mechanisms, achieving the first purely sense-level representation learning system with linear-time sense selection.We leverage reinforcement learning to enable joint training on the proposed modules, and introduce various exploration techniques on sense selection for better robustness.The experiments on benchmark data show that the proposed approach achieves the state-of-the-art performance on synonym selection as well as on contextual word similarities in terms of MaxSimC.

This paper emphasizes optimizing ranking measures in a recommendation problem. Since ranking measures are non-differentiable, previous works have been proposed to deal with this problem via approximations or lower/upper bounding of the loss. However, such mismatch between ranking measures and approximations/bounds can lead to non-optimal ranking results. To solve this problem, we propose to model the gradient of non-differentiable ranking measure based on the idea of virtual gradient, which is called lambda in learning to rank. In addition, noticing the difference between learning to rank and recommendation models, we prove that under certain circumstance the existence of popular items can lead to unlimited norm growing of the latent factors in a matrix factorization model. We further create a novel regularization term to remedy such concern. Finally, we demonstrate that our model, LambdaMF, outperforms several state-of-the-art methods. We further show in experiments that in all cases our model achieves global optimum of normalized discount cumulative gain during training. Detailed implementation and supplementary material can be found at (http://www.csie.ntu.edu.tw/~b00902055/).