Future activity anticipation is a challenging problem in egocentric vision. As a standard future activity anticipation paradigm, recursive sequence prediction suffers from the accumulation of errors. To address this problem, we propose a simple and effective Self-Regulated Learning framework, which aims to regulate the intermediate representation consecutively to produce representation that (a) emphasizes the novel information in the frame of the current time-stamp in contrast to previously observed content, and (b) reflects its correlation with previously observed frames. The former is achieved by minimizing a contrastive loss, and the latter can be achieved by a dynamic reweighing mechanism to attend to informative frames in the observed content with a similarity comparison between feature of the current frame and observed frames. The learned final video representation can be further enhanced by multi-task learning which performs joint feature learning on the target activity labels and the automatically detected action and object class tokens. SRL sharply outperforms existing state-of-the-art in most cases on two egocentric video datasets and two third-person video datasets. Its effectiveness is also verified by the experimental fact that the action and object concepts that support the activity semantics can be accurately identified.
Despite the recent success of multi-task learning and transfer learning for natural language processing (NLP), few works have systematically studied the effect of scaling up the number of tasks during pre-training. Towards this goal, this paper introduces ExMix (Extreme Mixture): a massive collection of 107 supervised NLP tasks across diverse domains and task-families. Using ExMix, we study the effect of multi-task pre-training at the largest scale to date, and analyze co-training transfer amongst common families of tasks. Through this analysis, we show that manually curating an ideal set of tasks for multi-task pre-training is not straightforward, and that multi-task scaling can vastly improve models on its own. Finally, we propose ExT5: a model pre-trained using a multi-task objective of self-supervised span denoising and supervised ExMix. Via extensive experiments, we show that ExT5 outperforms strong T5 baselines on SuperGLUE, GEM, Rainbow, Closed-Book QA tasks, and several tasks outside of ExMix. ExT5 also significantly improves sample efficiency while pre-training.
Multi-Task Learning (MTL) has achieved success in various fields. However, how to balance different tasks to achieve good performance is a key problem. To achieve the task balancing, there are many works to carefully design dynamical loss/gradient weighting strategies but the basic random experiments are ignored to examine their effectiveness. In this paper, we propose the Random Weighting (RW) methods, including Random Loss Weighting (RLW) and Random Gradient Weighting (RGW), where an MTL model is trained with random loss/gradient weights sampled from a distribution. To show the effectiveness and necessity of RW methods, theoretically we analyze the convergence of RW and reveal that RW has a higher probability to escape local minima, resulting in better generalization ability. Empirically, we extensively evaluate the proposed RW methods to compare with twelve state-of-the-art methods on five image datasets and two multilingual problems from the XTREME benchmark to show RW methods can achieve comparable performance with state-of-the-art baselines. Therefore, we think that the RW methods are important baselines for MTL and should attract more attentions.
In recent years, Multi-Task Learning (MTL) has attracted much attention due to its good performance in many applications. However, many existing MTL models cannot guarantee that their performance is no worse than their single-task counterparts on each task. Though some works have empirically observed this phenomenon, little work aims to handle the resulting problem. In this paper, we formally define this phenomenon as negative sharing and define safe multi-task learning where no negative sharing occurs. To achieve safe multi-task learning, we propose a Deep Safe Multi-Task Learning (DSMTL) model with two learning strategies: individual learning and joint learning. We theoretically study the safeness of both learning strategies in the DSMTL model to show that the proposed methods can achieve some versions of safe multi-task learning. Moreover, to improve the scalability of the DSMTL model, we propose an extension, which automatically learns a compact architecture and empirically achieves safe multi-task learning. Extensive experiments on benchmark datasets verify the safeness of the proposed methods.
Presentation attacks are recurrent threats to biometric systems, where impostors attempt to bypass these systems. Humans often use background information as contextual cues for their visual system. Yet, regarding face-based systems, the background is often discarded, since face presentation attack detection (PAD) models are mostly trained with face crops. This work presents a comparative study of face PAD models (including multi-task learning, adversarial training and dynamic frame selection) in two settings: with and without crops. The results show that the performance is consistently better when the background is present in the images. The proposed multi-task methodology beats the state-of-the-art results on the ROSE-Youtu dataset by a large margin with an equal error rate of 0.2%. Furthermore, we analyze the models' predictions with Grad-CAM++ with the aim to investigate to what extent the models focus on background elements that are known to be useful for human inspection. From this analysis we can conclude that the background cues are not relevant across all the attacks. Thus, showing the capability of the model to leverage the background information only when necessary.
Binaural audio provides human listeners with an immersive spatial sound experience, but most existing videos lack binaural audio recordings. We propose an audio spatialization method that draws on visual information in videos to convert their monaural (single-channel) audio to binaural audio. Whereas existing approaches leverage visual features extracted directly from video frames, our approach explicitly disentangles the geometric cues present in the visual stream to guide the learning process. In particular, we develop a multi-task framework that learns geometry-aware features for binaural audio generation by accounting for the underlying room impulse response, the visual stream's coherence with the sound source(s) positions, and the consistency in geometry of the sounding objects over time. Furthermore, we introduce a new large video dataset with realistic binaural audio simulated for real-world scanned environments. On two datasets, we demonstrate the efficacy of our method, which achieves state-of-the-art results.
In this paper, we propose a single UniFied transfOrmer (UFO), which is capable of processing either unimodal inputs (e.g., image or language) or multimodal inputs (e.g., the concatenation of the image and the question), for vision-language (VL) representation learning. Existing approaches typically design an individual network for each modality and/or a specific fusion network for multimodal tasks. To simplify the network architecture, we use a single transformer network and enforce multi-task learning during VL pre-training, which includes the image-text contrastive loss, image-text matching loss, and masked language modeling loss based on the bidirectional and the seq2seq attention mask. The same transformer network is used as the image encoder, the text encoder, or the fusion network in different pre-training tasks. Empirically, we observe less conflict among different tasks and achieve new state of the arts on visual question answering, COCO image captioning (cross-entropy optimization) and nocaps (in SPICE). On other downstream tasks, e.g., image-text retrieval, we also achieve competitive performance.
Machine learning shows great potential in virtual screening for drug discovery. Current efforts on accelerating docking-based virtual screening do not consider using existing data of other previously developed targets. To make use of the knowledge of the other targets and take advantage of the existing data, in this work, we apply multi-task learning to the problem of docking-based virtual screening. With two large docking datasets, the results of extensive experiments show that multi-task learning can achieve better performances on docking score prediction. By learning knowledge across multiple targets, the model trained by multi-task learning shows a better ability to adapt to a new target. Additional empirical study shows that other problems in drug discovery, such as the experimental drug-target affinity prediction, may also benefit from multi-task learning. Our results demonstrate that multi-task learning is a promising machine learning approach for docking-based virtual screening and accelerating the process of drug discovery.
In Federated Learning (FL), multiple clients collaborate to learn a shared model through a central server while keeping data decentralized. Personalized Federated Learning (PFL) further extends FL by learning a personalized model per client. In both FL and PFL, all clients participate in the training process and their labeled data are used for training. However, in reality, novel clients may wish to join a prediction service after it has been deployed, obtaining predictions for their own \textbf{unlabeled} data. Here, we introduce a new learning setup, On-Demand Unlabeled PFL (OD-PFL), where a system trained on a set of clients, needs to be later applied to novel unlabeled clients at inference time. We propose a novel approach to this problem, ODPFL-HN, which learns to produce a new model for the late-to-the-party client. Specifically, we train an encoder network that learns a representation for a client given its unlabeled data. That client representation is fed to a hypernetwork that generates a personalized model for that client. Evaluated on five benchmark datasets, we find that ODPFL-HN generalizes better than the current FL and PFL methods, especially when the novel client has a large shift from training clients. We also analyzed the generalization error for novel clients, and showed analytically and experimentally how novel clients can apply differential privacy.
Mispronunciation detection and diagnosis (MDD) is a popular research focus in computer-aided pronunciation training (CAPT) systems. End-to-end (e2e) approaches are becoming dominant in MDD. However an e2e MDD model usually requires entire speech utterances as input context, which leads to significant time latency especially for long paragraphs. We propose a streaming e2e MDD model called CoCA-MDD. We utilize conv-transformer structure to encode input speech in a streaming manner. A coupled cross-attention (CoCA) mechanism is proposed to integrate frame-level acoustic features with encoded reference linguistic features. CoCA also enables our model to perform mispronunciation classification with whole utterances. The proposed model allows system fusion between the streaming output and mispronunciation classification output for further performance enhancement. We evaluate CoCA-MDD on publicly available corpora. CoCA-MDD achieves F1 scores of 57.03% and 60.78% for streaming and fusion modes respectively on L2-ARCTIC. For phone-level pronunciation scoring, CoCA-MDD achieves 0.58 Pearson correlation coefficient (PCC) value on SpeechOcean762.
Shape information is crucial for human perception and cognition, and should therefore also play a role in cognitive AI systems. We employ the interdisciplinary framework of conceptual spaces, which proposes a geometric representation of conceptual knowledge through low-dimensional interpretable similarity spaces. These similarity spaces are often based on psychological dissimilarity ratings for a small set of stimuli, which are then transformed into a spatial representation by a technique called multidimensional scaling. Unfortunately, this approach is incapable of generalizing to novel stimuli. In this paper, we use convolutional neural networks to learn a generalizable mapping between perceptual inputs (pixels of grayscale line drawings) and a recently proposed psychological similarity space for the shape domain. We investigate different network architectures (classification network vs. autoencoder) and different training regimes (transfer learning vs. multi-task learning). Our results indicate that a classification-based multi-task learning scenario yields the best results, but that its performance is relatively sensitive to the dimensionality of the similarity space.
Recent advancements in end-to-end speech synthesis have made it possible to generate highly natural speech. However, training these models typically requires a large amount of high-fidelity speech data, and for unseen texts, the prosody of synthesized speech is relatively unnatural. To address these issues, we propose to combine a fine-tuned BERT-based front-end with a pre-trained FastSpeech2-based acoustic model to improve prosody modeling. The pre-trained BERT is fine-tuned on the polyphone disambiguation task, the joint Chinese word segmentation (CWS) and part-of-speech (POS) tagging task, and the prosody structure prediction (PSP) task in a multi-task learning framework. FastSpeech 2 is pre-trained on large-scale external data that are noisy but easier to obtain. Experimental results show that both the fine-tuned BERT model and the pre-trained FastSpeech 2 can improve prosody, especially for those structurally complex sentences.
It has been shown that machine translation models usually generate poor translations for named entities that are infrequent in the training corpus. Earlier named entity translation methods mainly focus on phonetic transliteration, which ignores the sentence context for translation and is limited in domain and language coverage. To address this limitation, we propose DEEP, a DEnoising Entity Pre-training method that leverages large amounts of monolingual data and a knowledge base to improve named entity translation accuracy within sentences. Besides, we investigate a multi-task learning strategy that finetunes a pre-trained neural machine translation model on both entity-augmented monolingual data and parallel data to further improve entity translation. Experimental results on three language pairs demonstrate that \method results in significant improvements over strong denoising auto-encoding baselines, with a gain of up to 1.3 BLEU and up to 9.2 entity accuracy points for English-Russian translation.
Multi-task learning is useful in NLP because it is often practically desirable to have a single model that works across a range of tasks. In the medical domain, sequential training on tasks may sometimes be the only way to train models, either because access to the original (potentially sensitive) data is no longer available, or simply owing to the computational costs inherent to joint retraining. A major issue inherent to sequential learning, however, is catastrophic forgetting, i.e., a substantial drop in accuracy on prior tasks when a model is updated for a new task. Elastic Weight Consolidation is a recently proposed method to address this issue, but scaling this approach to the modern large models used in practice requires making strong independence assumptions about model parameters, limiting its effectiveness. In this work, we apply Kronecker Factorization--a recent approach that relaxes independence assumptions--to prevent catastrophic forgetting in convolutional and Transformer-based neural networks at scale. We show the effectiveness of this technique on the important and illustrative task of medical entity linking across three datasets, demonstrating the capability of the technique to be used to make efficient updates to existing methods as new medical data becomes available. On average, the proposed method reduces catastrophic forgetting by 51% when using a BERT-based model, compared to a 27% reduction using standard Elastic Weight Consolidation, while maintaining spatial complexity proportional to the number of model parameters.
Nowadays, Knowledge graphs (KGs) have been playing a pivotal role in AI-related applications. Despite the large sizes, existing KGs are far from complete and comprehensive. In order to continuously enrich KGs, automatic knowledge construction and update mechanisms are usually utilized, which inevitably bring in plenty of noise. However, most existing knowledge graph embedding (KGE) methods assume that all the triple facts in KGs are correct, and project both entities and relations into a low-dimensional space without considering noise and knowledge conflicts. This will lead to low-quality and unreliable representations of KGs. To this end, in this paper, we propose a general multi-task reinforcement learning framework, which can greatly alleviate the noisy data problem. In our framework, we exploit reinforcement learning for choosing high-quality knowledge triples while filtering out the noisy ones. Also, in order to take full advantage of the correlations among semantically similar relations, the triple selection processes of similar relations are trained in a collective way with multi-task learning. Moreover, we extend popular KGE models TransE, DistMult, ConvE and RotatE with the proposed framework. Finally, the experimental validation shows that our approach is able to enhance existing KGE models and can provide more robust representations of KGs in noisy scenarios.
Neural processes have recently emerged as a class of powerful neural latent variable models that combine the strengths of neural networks and stochastic processes. As they can encode contextual data in the network's function space, they offer a new way to model task relatedness in multi-task learning. To study its potential, we develop multi-task neural processes, a new variant of neural processes for multi-task learning. In particular, we propose to explore transferable knowledge from related tasks in the function space to provide inductive bias for improving each individual task. To do so, we derive the function priors in a hierarchical Bayesian inference framework, which enables each task to incorporate the shared knowledge provided by related tasks into its context of the prediction function. Our multi-task neural processes methodologically expand the scope of vanilla neural processes and provide a new way of exploring task relatedness in function spaces for multi-task learning. The proposed multi-task neural processes are capable of learning multiple tasks with limited labeled data and in the presence of domain shift. We perform extensive experimental evaluations on several benchmarks for the multi-task regression and classification tasks. The results demonstrate the effectiveness of multi-task neural processes in transferring useful knowledge among tasks for multi-task learning and superior performance in multi-task classification and brain image segmentation.
Multi-task learning aims to explore task relatedness to improve individual tasks, which is of particular significance in the challenging scenario that only limited data is available for each task. To tackle this challenge, we propose variational multi-task learning (VMTL), a general probabilistic inference framework for learning multiple related tasks. We cast multi-task learning as a variational Bayesian inference problem, in which task relatedness is explored in a unified manner by specifying priors. To incorporate shared knowledge into each task, we design the prior of a task to be a learnable mixture of the variational posteriors of other related tasks, which is learned by the Gumbel-Softmax technique. In contrast to previous methods, our VMTL can exploit task relatedness for both representations and classifiers in a principled way by jointly inferring their posteriors. This enables individual tasks to fully leverage inductive biases provided by related tasks, therefore improving the overall performance of all tasks. Experimental results demonstrate that the proposed VMTL is able to effectively tackle a variety of challenging multi-task learning settings with limited training data for both classification and regression. Our method consistently surpasses previous methods, including strong Bayesian approaches, and achieves state-of-the-art performance on five benchmark datasets.
We propose a novel shape-aware relation network for accurate and real-time landmark detection in endoscopic submucosal dissection (ESD) surgery. This task is of great clinical significance but extremely challenging due to bleeding, lighting reflection, and motion blur in the complicated surgical environment. Compared with existing solutions, which either neglect geometric relationships among targeting objects or capture the relationships by using complicated aggregation schemes, the proposed network is capable of achieving satisfactory accuracy while maintaining real-time performance by taking full advantage of the spatial relations among landmarks. We first devise an algorithm to automatically generate relation keypoint heatmaps, which are able to intuitively represent the prior knowledge of spatial relations among landmarks without using any extra manual annotation efforts. We then develop two complementary regularization schemes to progressively incorporate the prior knowledge into the training process. While one scheme introduces pixel-level regularization by multi-task learning, the other integrates global-level regularization by harnessing a newly designed grouped consistency evaluator, which adds relation constraints to the proposed network in an adversarial manner. Both schemes are beneficial to the model in training, and can be readily unloaded in inference to achieve real-time detection. We establish a large in-house dataset of ESD surgery for esophageal cancer to validate the effectiveness of our proposed method. Extensive experimental results demonstrate that our approach outperforms state-of-the-art methods in terms of accuracy and efficiency, achieving better detection results faster. Promising results on two downstream applications further corroborate the great potential of our method in ESD clinical practice.
We propose a model-free reinforcement learning method for controlling mixed autonomy traffic in simulated traffic networks with through-traffic-only two-way and four-way intersections. Our method utilizes multi-agent policy decomposition which allows decentralized control based on local observations for an arbitrary number of controlled vehicles. We demonstrate that, even without reward shaping, reinforcement learning learns to coordinate the vehicles to exhibit traffic signal-like behaviors, achieving near-optimal throughput with 33-50% controlled vehicles. With the help of multi-task learning and transfer learning, we show that this behavior generalizes across inflow rates and size of the traffic network. Our code, models, and videos of results are available at https://github.com/ZhongxiaYan/mixed_autonomy_intersections.
Person search, which requires both pedestrian detection and person re-identification, is a challenging computer vision task applied to real-world scenarios. The challenges faced by detection and re-identification, such as occlusion, poor illumination, confusing background, are still urgent for person search. In addition, one-step methods for person search need to deal with the divergence between two tasks. In this work, we propose an end-to-end model containing the feature extractor, the region proposal network, and the multi-task learning module. In order to process divergence between detection and re-identification, we introduce switchable normalization and gradient centralization to improve the stability of the model. To solve the imbalance problem of hard examples, we introduce focal loss as a classification loss in the multi-task learning module. The experimental results on two bench-marks, i.e., CUHK-SYSU and PRW, well demonstrate that our method outperforms the state-of-the-art one-step methods.