Since its inception, the neural estimation of mutual information (MI) has demonstrated the empirical success of modeling expected dependency between high-dimensional random variables. However, MI is an aggregate statistic and cannot be used to measure point-wise dependency between different events. In this work, instead of estimating the expected dependency, we focus on estimating point-wise dependency (PD), which quantitatively measures how likely two outcomes co-occur. We show that we can naturally obtain PD when we are optimizing MI neural variational bounds. However, optimizing these bounds is challenging due to its large variance in practice. To address this issue, we develop two methods (free of optimizing MI variational bounds): Probabilistic Classifier and Density-Ratio Fitting. We demonstrate the effectiveness of our approaches in 1) MI estimation, 2) self-supervised representation learning, and 3) cross-modal retrieval task.
Self-supervised representation learning adopts self-defined signals as supervision and uses the learned representation for downstream tasks, such as masked language modeling (e.g., BERT) for natural language processing and contrastive visual representation learning (e.g., SimCLR) for computer vision applications. In this paper, we present a theoretical framework explaining that self-supervised learning is likely to work under the assumption that only the shared information (e.g., contextual information or content) between the input (e.g., non-masked words or original images) and self-supervised signals (e.g., masked-words or augmented images) contributes to downstream tasks. Under this assumption, we demonstrate that self-supervisedly learned representation can extract task-relevant and discard task-irrelevant information. We further connect our theoretical analysis to popular contrastive and predictive (self-supervised) learning objectives. In the experimental section, we provide controlled experiments on two popular tasks: 1) visual representation learning with various self-supervised learning objectives to empirically support our analysis; and 2) visual-textual representation learning to challenge that input and self-supervised signal lie in different modalities.
Aspect-based sentiment analysis (ABSA), a popular research area in NLP has two distinct parts -- aspect extraction (AE) and labeling the aspects with sentiment polarity (ALSA). Although distinct, these two tasks are highly correlated. The work primarily hypothesize that transferring knowledge from a pre-trained AE model can benefit the performance of ALSA models. Based on this hypothesis, word embeddings are obtained during AE and subsequently, feed that to the ALSA model. Empirically, this work show that the added information significantly improves the performance of three different baseline ALSA models on two distinct domains. This improvement also translates well across domains between AE and ALSA tasks.
The human language has heterogeneous sources of information, including tones of voice, facial gestures, and spoken language. Recent advances introduced computational models to combine these multimodal sources and yielded strong performance on human-centric tasks. Nevertheless, most of the models are often black-box, which comes with the price of lacking interpretability. In this paper, we propose Multimodal Routing to separate the contributions to the prediction from each modality and the interactions between modalities. At the heart of our method is a routing mechanism that represents each prediction as a concept, i.e., a vector in a Euclidean space. The concept assumes a linear aggregation from the contributions of multimodal features. Then, the routing procedure iteratively 1) associates a feature and a concept by checking how this concept agrees with this feature and 2) updates the concept based on the associations. In our experiments, we provide both global and local interpretation using Multimodal Routing on sentiment analysis and emotion prediction, without loss of performance compared to state-of-the-art methods. For example, we observe that our model relies mostly on the text modality for neutral sentiment predictions, the acoustic modality for extremely negative predictions, and the text-acoustic bimodal interaction for extremely positive predictions.
Multi-agent trajectory forecasting in autonomous driving requires an agent to accurately anticipate the behaviors of the surrounding vehicles and pedestrians, for safe and reliable decision-making. Due to partial observability over the goals, contexts, and interactions of agents in these dynamical scenes, directly obtaining the posterior distribution over future agent trajectories remains a challenging problem. In realistic embodied environments, each agent's future trajectories should be diverse since multiple plausible sequences of actions can be used to reach its intended goals, and they should be admissible since they must obey physical constraints and stay in drivable areas. In this paper, we propose a model that fully synthesizes multiple input signals from the multimodal world|the environment's scene context and interactions between multiple surrounding agents|to best model all diverse and admissible trajectories. We offer new metrics to evaluate the diversity of trajectory predictions, while ensuring admissibility of each trajectory. Based on our new metrics as well as those used in prior work, we compare our model with strong baselines and ablations across two datasets and show a 35% performance-improvement over the state-of-the-art.
Several recent works have found the emergence of grounded compositional language in the communication protocols developed by mostly cooperative multi-agent systems when learned end-to-end to maximize performance on a downstream task. However, human populations learn to solve complex tasks involving communicative behaviors not only in fully cooperative settings but also in scenarios where competition acts as an additional external pressure for improvement. In this work, we investigate whether competition for performance from an external, similar agent team could act as a social influence that encourages multi-agent populations to develop better communication protocols for improved performance, compositionality, and convergence speed. We start from Task & Talk, a previously proposed referential game between two cooperative agents as our testbed and extend it into Task, Talk & Compete, a game involving two competitive teams each consisting of two aforementioned cooperative agents. Using this new setting, we provide an empirical study demonstrating the impact of competitive influence on multi-agent teams. Our results show that an external competitive influence leads to improved accuracy and generalization, as well as faster emergence of communicative languages that are more informative and compositional.
Learning in the presence of label noise is a challenging yet important task: it is crucial to design models that are robust in the presence of mislabeled datasets. In this paper, we discover that a new class of loss functions called the gambler's loss provides strong robustness to label noise across various levels of corruption. We show that training with this loss function encourages the model to "abstain" from learning on the data points with noisy labels, resulting in a simple and effective method to improve robustness and generalization. In addition, we propose two practical extensions of the method: 1) an analytical early stopping criterion to approximately stop training before the memorization of noisy labels, as well as 2) a heuristic for setting hyperparameters which do not require knowledge of the noise corruption rate. We demonstrate the effectiveness of our method by achieving strong results across three image and text classification tasks as compared to existing baselines.
Federated learning is an emerging research paradigm to train models on private data distributed over multiple devices. A key challenge involves keeping private all the data on each device and training a global model only by communicating parameters and updates. Overcoming this problem relies on the global model being sufficiently compact so that the parameters can be efficiently sent over communication channels such as wireless internet. Given the recent trend towards building deeper and larger neural networks, deploying such models in federated settings on real-world tasks is becoming increasingly difficult. To this end, we propose to augment federated learning with local representation learning on each device to learn useful and compact features from raw data. As a result, the global model can be smaller since it only operates on higher-level local representations. We show that our proposed method achieves superior or competitive results when compared to traditional federated approaches on a suite of publicly available real-world datasets spanning image recognition (MNIST, CIFAR) and multimodal learning (VQA). Our choice of local representation learning also reduces the number of parameters and updates that need to be communicated to and from the global model, thereby reducing the bottleneck in terms of communication cost. Finally, we show that our local models provide flexibility in dealing with online heterogeneous data and can be easily modified to learn fair representations that obfuscate protected attributes such as race, age, and gender, a feature crucial to preserving the privacy of on-device data.
In recent years, extensive research has emerged in affective computing on topics like automatic emotion recognition and determining the signals that characterize individual emotions. Much less studied, however, is expressiveness, or the extent to which someone shows any feeling or emotion. Expressiveness is related to personality and mental health and plays a crucial role in social interaction. As such, the ability to automatically detect or predict expressiveness can facilitate significant advancements in areas ranging from psychiatric care to artificial social intelligence. Motivated by these potential applications, we present an extension of the BP4D+ dataset with human ratings of expressiveness and develop methods for (1) automatically predicting expressiveness from visual data and (2) defining relationships between interpretable visual signals and expressiveness. In addition, we study the emotional context in which expressiveness occurs and hypothesize that different sets of signals are indicative of expressiveness in different contexts (e.g., in response to surprise or in response to pain). Analysis of our statistical models confirms our hypothesis. Consequently, by looking at expressiveness separately in distinct emotional contexts, our predictive models show significant improvements over baselines and achieve comparable results to human performance in terms of correlation with the ground truth.