In Multi-Source Domain Adaptation (MSDA), models are trained on samples from multiple source domains and used for inference on a different, target, domain. Mainstream domain adaptation approaches learn a joint representation of source and target domains. Unfortunately, a joint representation may emphasize features that are useful for the source domains but hurt inference on target (negative transfer), or remove essential information about the target domain (knowledge fading). We propose Multi-source Student Teacher (MUST), a novel procedure designed to alleviate these issues. The key idea has two steps: First, we train a teacher network on source labels and infer pseudo labels on the target. Then, we train a student network using the pseudo labels and regularized the teacher to fit the student predictions. This regularization helps the teacher predictions on the target data remain consistent between epochs. Evaluations of MUST on three MSDA benchmarks: digits, text sentiment analysis, and visual-object recognition show that MUST outperforms current SoTA, sometimes by a very large margin. We further analyze the solutions and the dynamics of the optimization showing that the learned models follow the target distribution density, implicitly using it as information within the unlabeled target data.
Graph neural networks (GNNs) can process graphs of different sizes but their capacity to generalize across sizes is still not well understood. Size generalization is key to numerous GNN applications, from solving combinatorial optimization problems to learning in molecular biology. In such problems, obtaining labels and training on large graphs can be prohibitively expensive, but training on smaller graphs is possible. This paper puts forward the size-generalization question and characterizes important aspects of that problem theoretically and empirically. We show that even for very simple tasks, GNNs do not naturally generalize to graphs of larger size. Instead, their generalization performance is closely related to the distribution of patterns of connectivity and features and how that distribution changes from small to large graphs. Specifically, we show that in many cases, there are GNNs that can perfectly solve a task on small graphs but generalize poorly to large graphs and that these GNNs are encountered in practice. We then formalize size generalization as a domain-adaption problem and describe two learning setups where size generalization can be improved. First, as a self-supervised learning problem (SSL) over the target domain of large graphs. Second, as a semi-supervised learning problem when few samples are available in the target domain. We demonstrate the efficacy of these solutions on a diverse set of benchmark graph datasets.
We consider the problem of monitoring and controlling a partially-observed dynamic process that spreads over a graph. This problem naturally arises in contexts such as scheduling virus tests or quarantining individuals to curb a spreading epidemic; detecting fake news spreading on online networks by manually inspecting posted articles; and targeted marketing where the objective is to encourage the spread of a product. Curbing the spread and constraining the fraction of infected population becomes challenging when only a fraction of the population can be tested or quarantined. To address this challenge, we formulate this setup as a sequential decision problem over a graph. In face of an exponential state space, combinatorial action space and partial observability, we design RLGN, a novel tractable Reinforcement Learning (RL) scheme to prioritize which nodes should be tested, using Graph Neural Networks (GNNs) to rank the graph nodes. We evaluate this approach in three types of social-networks: community-structured, preferential attachment, and based on statistics from real cellular tracking. RLGN consistently outperforms all baselines in our experiments. It suggests that prioritizing tests using RL on temporal graphs can increase the number of healthy people by $25\%$ and contain the epidemic $30\%$ more often than supervised approaches and $2.5\times$ more often than non-learned baselines using the same resources.
Multi-objective optimization problems are prevalent in machine learning. These problems have a set of optimal solutions, called the Pareto front, where each point on the front represents a different trade-off between possibly conflicting objectives. Recent optimization algorithms can target a specific desired ray in loss space, but still face two grave limitations: (i) A separate model has to be trained for each point on the front; and (ii) The exact trade-off must be known prior to the optimization process. Here, we tackle the problem of learning the entire Pareto front, with the capability of selecting a desired operating point on the front after training. We call this new setup Pareto-Front Learning (PFL). We describe an approach to PFL implemented using HyperNetworks, which we term Pareto HyperNetworks (PHNs). PHN learns the entire Pareto front simultaneously using a single hypernetwork, which receives as input a desired preference vector and returns a Pareto-optimal model whose loss vector is in the desired ray. The unified model is runtime efficient compared to training multiple models, and generalizes to new operating points not used during training. We evaluate our method on a wide set of problems, from multi-task regression and classification to fairness. PHNs learns the entire Pareto front in roughly the same time as learning a single point on the front, and also reaches a better solution set. PFL opens the door to new applications where models are selected based on preferences that are only available at run time.
We study the problem of recognizing visual entities from the textual descriptions of their classes. Specifically, given birds' images with free-text descriptions of their species, we learn to classify images of previously-unseen species based on specie descriptions. This setup has been studied in the vision community under the name zero-shot learning from text, focusing on learning to transfer knowledge about visual aspects of birds from seen classes to previously-unseen ones. Here, we suggest focusing on the textual description and distilling from the description the most relevant information to effectively match visual features to the parts of the text that discuss them. Specifically, (1) we propose to leverage the similarity between species, reflected in the similarity between text descriptions of the species. (2) we derive visual summaries of the texts, i.e., extractive summaries that focus on the visual features that tend to be reflected in images. We propose a simple attention-based model augmented with the similarity and visual summaries components. Our empirical results consistently and significantly outperform the state-of-the-art on the largest benchmarks for text-based zero-shot learning, illustrating the critical importance of texts for zero-shot image-recognition.
Object detection is a fundamental task in computer vision, requiring large annotated datasets that are difficult to collect, as annotators need to label objects and their bounding boxes. Thus, it is a significant challenge to use cheaper forms of supervision effectively. Recent work has begun to explore image captions as a source for weak supervision, but to date, in the context of object detection, captions have only been used to infer the categories of the objects in the image. In this work, we argue that captions contain much richer information about the image, including attributes of objects and their relations. Namely, the text represents a scene of the image, as described recently in the literature. We present a method that uses the attributes in this "textual scene graph" to train object detectors. We empirically demonstrate that the resulting model achieves state-of-the-art results on several challenging object detection datasets, outperforming recent approaches.
Videos of actions are complex spatio-temporal signals, containing rich compositional structures. Current generative models are limited in their ability to generate examples of object configurations outside the range they were trained on. Towards this end, we introduce a generative model (AG2Vid) based on Action Graphs, a natural and convenient structure that represents the dynamics of actions between objects over time. Our AG2Vid model disentangles appearance and position features, allowing for more accurate generation. AG2Vid is evaluated on the CATER and Something-Something datasets and outperforms other baselines. Finally, we show how Action Graphs can be used for generating novel compositions of unseen actions.
People easily recognize new visual categories that are new combinations of known components. This compositional generalization capacity is critical for learning in real-world domains like vision and language because the long tail of new combinations dominates the distribution. Unfortunately, learning systems struggle with compositional generalization because they often build on features that are correlated with class labels even if they are not "essential" for the class. This leads to consistent misclassification of samples from a new distribution, like new combinations of known components. Here we describe an approach for compositional generalization that builds on causal ideas. First, we describe compositional zero-shot learning from a causal perspective, and propose to view zero-shot inference as finding "which intervention caused the image?". Second, we present a causal-inspired embedding model that learns disentangled representations of elementary components of visual objects from correlated (confounded) training data. We evaluate this approach on two datasets for predicting new combinations of attribute-object pairs: A well-controlled synthesized images dataset and a real world dataset which consists of fine-grained types of shoes. We show improvements compared to strong baselines.
Training with multiple auxiliary tasks is a common practice used in deep learning for improving the performance on the main task of interest. Two main challenges arise in this multi-task learning setting: (i) Designing useful auxiliary tasks; and (ii) Combining auxiliary tasks into a single coherent loss. We propose a novel framework, \textit{AuxiLearn}, that targets both challenges, based on implicit differentiation. First, when useful auxiliaries are known, we propose learning a network that combines all losses into a single coherent objective function. This network can learn \textit{non-linear} interactions between auxiliary tasks. Second, when no useful auxiliary task is known, we describe how to learn a network that generates a meaningful, novel auxiliary task. We evaluate AuxiLearn in a series of tasks and domains, including image segmentation and learning with attributes. We find that AuxiLearn consistently improves accuracy compared with competing methods.