The field of meta-learning seeks to improve the ability of today's machine learning systems to adapt efficiently to small amounts of data. Typically this is accomplished by training a system with a parametrized update rule to improve a task-relevant objective based on supervision or a reward function. However, in many domains of practical interest, task data is unlabeled, or reward functions are unavailable. In this paper we introduce a new approach to address the more general problem of generative meta-learning, which we argue is an important prerequisite for obtaining human-level cognitive flexibility in artificial agents, and can benefit many practical applications along the way. Our contribution leverages the AEVB framework and mean-field variational Bayes, and creates fast-adapting latent-space generative models. At the heart of our contribution is a new result, showing that for a broad class of deep generative latent variable models, the relevant VB updates do not depend on any generative neural network. The theoretical merits of our approach are reflected in empirical experiments.

Transfer learning across heterogeneous data distributions (a.k.a. domains) and distinct tasks is a more general and challenging problem than conventional transfer learning, where either domains or tasks are assumed to be the same. While neural network based feature transfer is widely used in transfer learning applications, finding the optimal transfer strategy still requires time-consuming experiments and domain knowledge. We propose a transferability metric called Optimal Transport based Conditional Entropy (OTCE), to analytically predict the transfer performance for supervised classification tasks in such cross-domain and cross-task feature transfer settings. Our OTCE score characterizes transferability as a combination of domain difference and task difference, and explicitly evaluates them from data in a unified framework. Specifically, we use optimal transport to estimate domain difference and the optimal coupling between source and target distributions, which is then used to derive the conditional entropy of the target task (task difference). Experiments on the largest cross-domain dataset DomainNet and Office31 demonstrate that OTCE shows an average of 21% gain in the correlation with the ground truth transfer accuracy compared to state-of-the-art methods. We also investigate two applications of the OTCE score including source model selection and multi-source feature fusion.

We introduce the problem of robust subgroup discovery, i.e., finding a set of interpretable descriptions of subsets that 1) stand out with respect to one or more target attributes, 2) are statistically robust, and 3) non-redundant. Many attempts have been made to mine either locally robust subgroups or to tackle the pattern explosion, but we are the first to address both challenges at the same time from a global perspective. First, we formulate a broad model class of subgroup lists, i.e., ordered sets of subgroups, for univariate and multivariate targets that can consist of nominal or numeric variables. This novel model class allows us to formalize the problem of optimal robust subgroup discovery using the Minimum Description Length (MDL) principle, where we resort to optimal Normalized Maximum Likelihood and Bayesian encodings for nominal and numeric targets, respectively. Notably, we show that our problem definition is equal to mining the top-1 subgroup with an information-theoretic quality measure plus a penalty for complexity. Second, as finding optimal subgroup lists is NP-hard, we propose RSD, a greedy heuristic that finds good subgroup lists and guarantees that the most significant subgroup found according to the MDL criterion is added in each iteration, which is shown to be equivalent to a Bayesian one-sample proportions, multinomial, or t-test between the subgroup and dataset marginal target distributions plus a multiple hypothesis testing penalty. We empirically show on 54 datasets that RSD outperforms previous subgroup set discovery methods in terms of quality and subgroup list size.

Class labels used for machine learning are relatable to each other, with certain class labels being more similar to each other than others (e.g. images of cats and dogs are more similar to each other than those of cats and cars). Such similarity among classes is often the cause of poor model performance due to the models confusing between them. Current labeling techniques fail to explicitly capture such similarity information. In this paper, we instead exploit the similarity between classes by capturing the similarity information with our novel confidence labels. Confidence labels are probabilistic labels denoting the likelihood of similarity, or confusability, between the classes. Often even after models are trained to differentiate between classes in the feature space, the similar classes' latent space still remains clustered. We view this type of clustering as valuable information and exploit it with our novel projective loss functions. Our projective loss functions are designed to work with confidence labels with an ability to relax the loss penalty for errors that confuse similar classes. We use our approach to train neural networks with noisy labels, as we believe noisy labels are partly a result of confusability arising from class similarity. We show improved performance compared to the use of standard loss functions. We conduct a detailed analysis using the CIFAR-10 dataset and show our proposed methods' applicability to larger datasets, such as ImageNet and Food-101N.

Action segmentation refers to inferring boundaries of semantically consistent visual concepts in videos and is an important requirement for many video understanding tasks. For this and other video understanding tasks, supervised approaches have achieved encouraging performance but require a high volume of detailed frame-level annotations. We present a fully automatic and unsupervised approach for segmenting actions in a video that does not require any training. Our proposal is an effective temporally-weighted hierarchical clustering algorithm that can group semantically consistent frames of the video. Our main finding is that representing a video with a 1-nearest neighbor graph by taking into account the time progression is sufficient to form semantically and temporally consistent clusters of frames where each cluster may represent some action in the video. Additionally, we establish strong unsupervised baselines for action segmentation and show significant performance improvements over published unsupervised methods on five challenging action segmentation datasets. Our approach also outperforms weakly-supervised methods by large margins on 4 of these datasets. Interestingly, we also achieve better results than many fully-supervised methods that have reported results on these datasets. Our code is available at https://github.com/ssarfraz/FINCH-Clustering/tree/master/TW-FINCH

Label noise in multiclass classification is a major obstacle to the deployment of learning systems. However, unlike the widely used class-conditional noise (CCN) assumption that the noisy label is independent of the input feature given the true label, label noise in real-world datasets can be aleatory and heavily dependent on individual instances. In this work, we investigate the instance-dependent noise (IDN) model and propose an efficient approximation of IDN to capture the instance-specific label corruption. Concretely, noting the fact that most columns of the IDN transition matrix have only limited influence on the class-posterior estimation, we propose a variational approximation that uses a single-scalar confidence parameter. To cope with the situation where the mapping from the instance to its confidence value could vary significantly for two adjacent instances, we suggest using instance embedding that assigns a trainable parameter to each instance. The resulting instance-confidence embedding (ICE) method not only performs well under label noise but also can effectively detect ambiguous or mislabeled instances. We validate its utility on various image and text classification tasks.

In a variety of settings, limitations of sensing technologies or other sampling mechanisms result in missing labels, where the likelihood of a missing label in the training set is an unknown function of the data. For example, satellites used to detect forest fires cannot sense fires below a certain size threshold. In such cases, training datasets consist of positive and pseudo-negative observations where pseudo-negative observations can be either true negatives or undetected positives with small magnitudes. We develop a new methodology and non-convex algorithm P(ositive) U(nlabeled) - O(ccurrence) M(agnitude) M(ixture) which jointly estimates the occurrence and detection likelihood of positive samples, utilizing prior knowledge of the detection mechanism. Our approach uses ideas from positive-unlabeled (PU)-learning and zero-inflated models that jointly estimate the magnitude and occurrence of events. We provide conditions under which our model is identifiable and prove that even though our approach leads to a non-convex objective, any local minimizer has optimal statistical error (up to a log term) and projected gradient descent has geometric convergence rates. We demonstrate on both synthetic data and a California wildfire dataset that our method out-performs existing state-of-the-art approaches.

Non-convex optimization is ubiquitous in modern machine learning. Researchers devise non-convex objective functions and optimize them using off-the-shelf optimizers such as stochastic gradient descent and its variants, which leverage the local geometry and update iteratively. Even though solving non-convex functions is NP-hard in the worst case, the optimization quality in practice is often not an issue -- optimizers are largely believed to find approximate global minima. Researchers hypothesize a unified explanation for this intriguing phenomenon: most of the local minima of the practically-used objectives are approximately global minima. We rigorously formalize it for concrete instances of machine learning problems.

When fitting statistical models, some predictors are often found to be correlated with each other, and functioning together. Many group variable selection methods are developed to select the groups of predictors that are closely related to the continuous or categorical response. These existing methods usually assume the group structures are well known. For example, variables with similar practical meaning, or dummy variables created by categorical data. However, in practice, it is impractical to know the exact group structure, especially when the variable dimensional is large. As a result, the group variable selection results may be selected. To solve the challenge, we propose a two-stage approach that combines a variable clustering stage and a group variable stage for the group variable selection problem. The variable clustering stage uses information from the data to find a group structure, which improves the performance of the existing group variable selection methods. For ultrahigh dimensional data, where the predictors are much larger than observations, we incorporated a variable screening method in the first stage and shows the advantages of such an approach. In this article, we compared and discussed the performance of four existing group variable selection methods under different simulation models, with and without the variable clustering stage. The two-stage method shows a better performance, in terms of the prediction accuracy, as well as in the accuracy to select active predictors. An athlete's data is also used to show the advantages of the proposed method.

Joint probability mass function (PMF) estimation is a fundamental machine learning problem. The number of free parameters scales exponentially with respect to the number of random variables. Hence, most work on nonparametric PMF estimation is based on some structural assumptions such as clique factorization adopted by probabilistic graphical models, imposition of low rank on the joint probability tensor and reconstruction from 3-way or 2-way marginals, etc. In the present work, we link random projections of data to the problem of PMF estimation using ideas from tomography. We integrate this idea with the idea of low-rank tensor decomposition to show that we can estimate the joint density from just one-way marginals in a transformed space. We provide a novel algorithm for recovering factors of the tensor from one-way marginals, test it across a variety of synthetic and real-world datasets, and also perform MAP inference on the estimated model for classification.