Self-supervised learning is a powerful way to learn useful representations from natural data. It has also been suggested as one possible means of building visual representation in humans, but the specific objective and algorithm are unknown. Currently, most self-supervised methods encourage the system to learn an invariant representation of different transformations of the same image in contrast to those of other images. However, such transformations are generally non-biologically plausible, and often consist of contrived perceptual schemes such as random cropping and color jittering. In this paper, we attempt to reverse-engineer these augmentations to be more biologically or perceptually plausible while still conferring the same benefits for encouraging robust representation. Critically, we find that random cropping can be substituted by cortical magnification, and saccade-like sampling of the image could also assist the representation learning. The feasibility of these transformations suggests a potential way that biological visual systems could implement self-supervision. Further, they break the widely accepted spatially-uniform processing assumption used in many computer vision algorithms, suggesting a role for spatially-adaptive computation in humans and machines alike. Our code and demo can be found here (Wang, 2021).

Prediction using the ground truth sounds like an oxymoron in machine learning. However, such an unrealistic setting was used in hundreds, if not thousands of papers in the area of finding graph representations. To evaluate the multi-label problem of node classification by using the obtained representations, many works assume in the prediction stage that the number of labels of each test instance is known. In practice such ground truth information is rarely available, but we point out that such an inappropriate setting is now ubiquitous in this research area. We detailedly investigate why the situation occurs. Our analysis indicates that with unrealistic information, the performance is likely over-estimated. To see why suitable predictions were not used, we identify difficulties in applying some multi-label techniques. For the use in future studies, we propose simple and effective settings without using practically unknown information. Finally, we take this chance to conduct a fair and serious comparison of major graph-representation learning methods on multi-label node classification.

Natural Language Interfaces to Databases (NLIDBs), where users pose queries in Natural Language (NL), are crucial for enabling non-experts to gain insights from data. Developing such interfaces, by contrast, is dependent on experts who often code heuristics for mapping NL to SQL. Alternatively, NLIDBs based on machine learning models rely on supervised examples of NL to SQL mappings (NL-SQL pairs) used as training data. Such examples are again procured using experts, which typically involves more than a one-off interaction. Namely, each data domain in which the NLIDB is deployed may have different characteristics and therefore require either dedicated heuristics or domain-specific training examples. To this end, we propose an alternative approach for training machine learning-based NLIDBs, using weak supervision. We use the recently proposed question decomposition representation called QDMR, an intermediate between NL and formal query languages. Recent work has shown that non-experts are generally successful in translating NL to QDMR. We consequently use NL-QDMR pairs, along with the question answers, as supervision for automatically synthesizing SQL queries. The NL questions and synthesized SQL are then used to train NL-to-SQL models, which we test on five benchmark datasets. Extensive experiments show that our solution, requiring zero expert annotations, performs competitively with models trained on expert annotated data.

The performance of supervised learning tasks improves with more high-quality labels available. However, it is expensive to collect a large number of labeled ...

In this paper, we approach the problem of semantic search by framing the search task as paraphrase span detection, i.e. given a segment of text as a query phrase, the task is to identify its paraphrase in a given document, the same modelling setup as typically used in extractive question answering. On the Turku Paraphrase Corpus of 100,000 manually extracted Finnish paraphrase pairs including their original document context, we find that our paraphrase span detection model outperforms two strong retrieval baselines (lexical similarity and BERT sentence embeddings) by 31.9pp and 22.4pp respectively in terms of exact match, and by 22.3pp and 12.9pp in terms of token-level F-score. This demonstrates a strong advantage of modelling the task in terms of span retrieval, rather than sentence similarity. Additionally, we introduce a method for creating artificial paraphrase data through back-translation, suitable for languages where manually annotated paraphrase resources for training the span detection model are not available.

Moderating content on today's internet is akin to a round of Whack-A-Mole with human moderators continually forced to react in realtime to changing trends, such as vaccine mis- and disinformation or intentional bad actors probing for ways around established personal conduct policies. Machine learning systems can help alleviate some of this burden by automating the policy enforcement process, however modern AI systems often require months of lead time to properly train and deploy (time mostly spent collecting and annotating the thousands, if not millions of, necessary examples). To shorten that response time, at least to a matter of weeks rather than months, Meta's AI research group (formerly FAIR) has developed a more generalized technology that requires just a handful of specific examples in order to respond to new and emerging forms of malicious content, called Few-Shot Learner (FSL). Few-shot learning is a relatively recent development in AI, essentially teaching the system to make accurate predictions based on a limited number of training examples -- quite the opposite of conventional supervised learning methods. For example, if you wanted to train a standard SL model to recognize pictures of rabbits, you feed it a couple hundred thousands of rabbit pictures and then you can present it with two images and ask if they both show the same animal. Thing is, the model doesn't know if the two pictures are of rabbits because it doesn't actually know what a rabbit is.

High levels of missing data and strong class imbalance are ubiquitous challenges that are often presented simultaneously in real-world time series data. Existing methods approach these problems separately, frequently making significant assumptions about the underlying data generation process in order to lessen the impact of missing information. In this work, we instead demonstrate how a general self-supervised training method, namely Autoregressive Predictive Coding (APC), can be leveraged to overcome both missing data and class imbalance simultaneously without strong assumptions. Specifically, on a synthetic dataset, we show that standard baselines are substantially improved upon through the use of APC, yielding the greatest gains in the combined setting of high missingness and severe class imbalance. We further apply APC on two real-world medical time-series datasets, and show that APC improves the classification performance in all settings, ultimately achieving state-of-the-art AUPRC results on the Physionet benchmark.

Column Generation (CG) is an effective method for solving large-scale optimization problems. CG starts by solving a sub-problem with a subset of columns (i.e., variables) and gradually includes new columns that can improve the solution of the current subproblem. The new columns are generated as needed by repeatedly solving a pricing problem, which is often NP-hard and is a bottleneck of the CG approach. To tackle this, we propose a Machine-Learning-based Pricing Heuristic (MLPH)that can generate many high-quality columns efficiently. In each iteration of CG, our MLPH leverages an ML model to predict the optimal solution of the pricing problem, which is then used to guide a sampling method to efficiently generate multiple high-quality columns. Using the graph coloring problem, we empirically show that MLPH significantly enhancesCG as compared to six state-of-the-art methods, and the improvement in CG can lead to substantially better performance of the branch-and-price exact method.

In the last years predict-and-optimize approaches (Elmachtoub and Grigas 2021; Wilder, Dilkina, and Tambe 2019) have received increasing attention. These problems have the settings where the predictions of predictive machine learning (ML) models are fed to downstream optimization problems for decision making. Predict-and-optimize approaches propose to train the ML models, often neural network models, by directly optimizing the quality of decisions made by the optimization solvers. However, one major bottleneck of predict-and-optimize approaches is solving the optimization problem for each training instance at every epoch. To address this challenge, Mulamba et al. (2021) propose noise contrastive estimation by caching feasible solutions. In this work, we show the noise contrastive estimation can be considered a case of learning to rank the solution cache. We also develop pairwise and listwise ranking loss functions, which can be differentiated in closed form without the need of solving the optimization problem. By training with respect to these surrogate loss function, we empirically show that we are able to minimize the regret of the predictions.

While modern policy optimization methods can do complex manipulation from sensory data, they struggle on problems with extended time horizons and multiple sub-goals. On the other hand, task and motion planning (TAMP) methods scale to long horizons but they are computationally expensive and need to precisely track world state. We propose a method that draws on the strength of both methods: we train a policy to imitate a TAMP solver's output. This produces a feed-forward policy that can accomplish multi-step tasks from sensory data. First, we build an asynchronous distributed TAMP solver that can produce supervision data fast enough for imitation learning. Then, we propose a hierarchical policy architecture that lets us use partially trained control policies to speed up the TAMP solver. In robotic manipulation tasks with 7-DoF joint control, the partially trained policies reduce the time needed for planning by a factor of up to 2.6. Among these tasks, we can learn a policy that solves the RoboSuite 4-object pick-place task 88% of the time from object pose observations and a policy that solves the RoboDesk 9-goal benchmark 79% of the time from RGB images (averaged across the 9 disparate tasks).