Commonsense knowledge graph (CKG) is a special type of knowledge graph (KG), where entities are composed of free-form text. However, most existing CKG completion methods focus on the setting where all the entities are presented at training time. Although this setting is standard for conventional KG completion, it has limitations for CKG completion. At test time, entities in CKGs can be unseen because they may have unseen text/names and entities may be disconnected from the training graph, since CKGs are generally very sparse. Here, we propose to study the inductive learning setting for CKG completion where unseen entities may present at test time. We develop a novel learning framework named InductivE. Different from previous approaches, InductiveE ensures the inductive learning capability by directly computing entity embeddings from raw entity attributes/text. InductiveE consists of a free-text encoder, a graph encoder, and a KG completion decoder. Specifically, the free-text encoder first extracts the textual representation of each entity based on the pre-trained language model and word embedding. The graph encoder is a gated relational graph convolutional neural network that learns from a densified graph for more informative entity representation learning. We develop a method that densifies CKGs by adding edges among semantic-related entities and provide more supportive information for unseen entities, leading to better generalization ability of entity embedding for unseen entities. Finally, inductiveE employs Conv-TransE as the CKG completion decoder. Experimental results show that InductiveE significantly outperforms state-of-the-art baselines in both standard and inductive settings on ATOMIC and ConceptNet benchmarks. InductivE performs especially well on inductive scenarios where it achieves above 48% improvement over present methods.

In analogy to compressed sensing, which allows sample-efficient signal reconstruction given prior knowledge of its sparsity in frequency domain, we propose to utilize policy simplicity (Occam's Razor) as a prior to enable sample-efficient imitation learning. We first demonstrated the feasibility of this scheme on linear case where state-value function can be sampled directly. We also extended the scheme to scenarios where only actions are visible and scenarios where the policy is obtained from nonlinear network. The method is benchmarked against behavior cloning and results in significantly higher scores with limited expert demonstrations.

Autoencoders are an unsupervised learning technique, although they are trained using supervised learning methods. The goal is to minimize reconstruction error based on a loss function, such as the mean squared error.

We considered a novel practical problem of online learning with episodically revealed rewards, motivated by several real-world applications, where the contexts are nonstationary over different episodes and the reward feedbacks are not always available to the decision making agents. For this online semi-supervised learning setting, we introduced Background Episodic Reward LinUCB (BerlinUCB), a solution that easily incorporates clustering as a self-supervision module to provide useful side information when rewards are not observed. Our experiments on a variety of datasets, both in stationary and nonstationary environments of six different scenarios, demonstrated clear advantages of the proposed approach over the standard contextual bandit. Lastly, we introduced a relevant real-life example where this problem setting is especially useful.

Distant supervision has been demonstrated to be highly beneficial to enhance relation extraction models, but it often suffers from high label noise. In this work, we propose a novel model-agnostic instance subsampling method for distantly supervised relation extraction, namely REIF, which bridges the gap of realizing influence subsampling in deep learning. It encompasses two key steps: first calculating instance-level influences that measure how much each training instance contributes to the validation loss change of our model, then deriving sampling probabilities via the proposed sigmoid sampling function to perform batch-in-bag sampling. We design a fast influence subsampling scheme that reduces the computational complexity from O(mn) to O(1), and analyze its robustness when the sigmoid sampling function is employed. Empirical experiments demonstrate our method's superiority over the baselines, and its ability to support interpretable instance selection.

The F-measure is a widely used performance measure for multi-label classification, where multiple labels can be active in an instance simultaneously (e.g. in image tagging, multiple tags can be active in any image). In particular, the F-measure explicitly balances recall (fraction of active labels predicted to be active) and precision (fraction of labels predicted to be active that are actually so), both of which are important in evaluating the overall performance of a multi-label classifier. As with most discrete prediction problems, however, directly optimizing the F-measure is computationally hard. In this paper, we explore the question of designing convex surrogate losses that are calibrated for the F-measure -- specifically, that have the property that minimizing the surrogate loss yields (in the limit of sufficient data) a Bayes optimal multi-label classifier for the F-measure. We show that the F-measure for an $s$-label problem, when viewed as a $2^s \times 2^s$ loss matrix, has rank at most $s^2+1$, and apply a result of Ramaswamy et al. (2014) to design a family of convex calibrated surrogates for the F-measure. The resulting surrogate risk minimization algorithms can be viewed as decomposing the multi-label F-measure learning problem into $s^2+1$ binary class probability estimation problems. We also provide a quantitative regret transfer bound for our surrogates, which allows any regret guarantees for the binary problems to be transferred to regret guarantees for the overall F-measure problem, and discuss a connection with the algorithm of Dembczynski et al. (2013). Our experiments confirm our theoretical findings.

Curation of large fully supervised datasets has become one of the major roadblocks for machine learning. Weak supervision provides an alternative to supervised learning by training with cheap, noisy, and possibly correlated labeling functions from varying sources. The key challenge in weakly supervised learning is combining the different weak supervision signals while navigating misleading correlations in their errors. In this paper, we propose a simple data-free approach for combining weak supervision signals by defining a constrained space for the possible labels of the weak signals and training with a random labeling within this constrained space. Our method is efficient and stable, converging after a few iterations of gradient descent. We prove theoretical conditions under which the worst-case error of the randomized label decreases with the rank of the linear constraints. We show experimentally that our method outperforms other weak supervision methods on various text- and image-classification tasks.

We proposed a novel AI framework to conduct real-time multi-speaker diarization and recognition without prior registration and pretraining in a fully online learning setting. Our contributions are two-fold. First, we proposed a new benchmark to evaluate the rarely studied fully online speaker diarization problem. We built upon existing datasets of real world utterances to automatically curate MiniVox, an experimental environment which generates infinite configurations of continuous multi-speaker speech stream. Secondly, we considered the practical problem of online learning with episodically revealed rewards and introduced a solution based on semi-supervised and self-supervised learning methods. Lastly, we provided a workable web-based recognition system which interactively handles the cold start problem of new user's addition by transferring representations of old arms to new ones with an extendable contextual bandit. We demonstrated that our proposed method obtained robust performance in the online MiniVox framework.

We present a new meta-learning approach for supervised anomaly classification / one-class classification using set equivariant networks. We focus our experiments on an astronomy application. Our problem setting is composed of a set of classification tasks. Each task has a (small) set of positive, labeled examples and a larger set of unlabeled examples. We expect the positive instances to be much more uncommon (i.e. 'anomalies') than the negative ones ('normal' class). We propose a novel use of equivariant networks for this setting. Specifically we use Deep Sets, which was developed for point-clouds and unordered sets and is equivariant to permutation. We propose to consider the set of positive examples of a given task as a 'point-cloud'. The key idea is that the network directly takes as input the set of positive examples in addition to the current example to classify. This allows the model to predict at test-time on new tasks using only positive labeled examples (i.e 'One-Class classification' setting) by design, potentially without retraining. However, the model is trained in a meta-learning regime on a dataset of several tasks with full-supervision (positive and negative labels). This setup is motivated by our target application on stellar streams. Streams are groups of stars sharing specific properties in various features. For a detected stream, we can determine a set of stars that likely belong to the stream. We aim to characterize the membership of all other nearby stars. We build a meta-dataset of simulated streams injected onto real data and evaluate on unseen synthetic streams and one known stream. Our experiments show encouraging results to explore furthermore equivariant networks for anomaly or 'one-class' classification in a meta-learning regime.

Fairness in machine learning has received considerable attention. However, most studies on fair learning focus on either supervised learning or unsupervised learning. Very few consider semi-supervised settings. Yet, in reality, most machine learning tasks rely on large datasets that contain both labeled and unlabeled data. One of key issues with fair learning is the balance between fairness and accuracy. Previous studies arguing that increasing the size of the training set can have a better trade-off. We believe that increasing the training set with unlabeled data may achieve the similar result. Hence, we develop a framework for fair semi-supervised learning, which is formulated as an optimization problem. This includes classifier loss to optimize accuracy, label propagation loss to optimize unlabled data prediction, and fairness constraints over labeled and unlabeled data to optimize the fairness level. The framework is conducted in logistic regression and support vector machines under the fairness metrics of disparate impact and disparate mistreatment. We theoretically analyze the source of discrimination in semi-supervised learning via bias, variance and noise decomposition. Extensive experiments show that our method is able to achieve fair semi-supervised learning, and reach a better trade-off between accuracy and fairness than fair supervised learning.