Neural language models (LMs) have shown to benefit significantly from enhancing word vectors with subword-level information, especially for morphologically rich languages. This has been mainly tackled by providing subword-level information as an input; using subword units in the output layer has been far less explored. In this work, we propose LMs that are cognizant of the underlying stems in each word. We derive stems for words using a simple unsupervised technique for stem identification. We experiment with different architectures involving multi-task learning and mixture models over words and stems. We focus on four morphologically complex languages -- Hindi, Tamil, Kannada and Finnish -- and observe significant perplexity gains with using our stem-driven LMs when compared with other competitive baseline models.
Meta-reinforcement learning algorithms can enable robots to acquire new skills much more quickly, by leveraging prior experience to learn how to learn. However, much of the current research on meta-reinforcement learning focuses on task distributions that are very narrow. For example, a commonly used meta-reinforcement learning benchmark uses different running velocities for a simulated robot as different tasks. When policies are meta-trained on such narrow task distributions, they cannot possibly generalize to more quickly acquire entirely new tasks. Therefore, if the aim of these methods is to enable faster acquisition of entirely new behaviors, we must evaluate them on task distributions that are sufficiently broad to enable generalization to new behaviors. In this paper, we propose an open-source simulated benchmark for meta-reinforcement learning and multi-task learning consisting of 50 distinct robotic manipulation tasks. Our aim is to make it possible to develop algorithms that generalize to accelerate the acquisition of entirely new, held-out tasks. We evaluate 7 state-of-the-art meta-reinforcement learning and multi-task learning algorithms on these tasks. Surprisingly, while each task and its variations (e.g., with different object positions) can be learned with reasonable success, these algorithms struggle to learn with multiple tasks at the same time, even with as few as ten distinct training tasks. Our analysis and open-source environments pave the way for future research in multi-task learning and meta-learning that can enable meaningful generalization, thereby unlocking the full potential of these methods.
The utility of linguistic annotation in neural machine translation seemed to had been established in past papers. The experiments were however limited to recurrent sequence-to-sequence architectures and relatively small data settings. We focus on the state-of-the-art Transformer model and use comparably larger corpora. Specifically, we try to promote the knowledge of source-side syntax using multi-task learning either through simple data manipulation techniques or through a dedicated model component. In particular, we train one of Transformer attention heads to produce source-side dependency tree. Overall, our results cast some doubt on the utility of multi-task setups with linguistic information. The data manipulation techniques, recommended in previous works, prove ineffective in large data settings. The treatment of self-attention as dependencies seems much more promising: it helps in translation and reveals that Transformer model can very easily grasp the syntactic structure. An important but curious result is, however, that identical gains are obtained by using trivial "linear trees" instead of true dependencies. The reason for the gain thus may not be coming from the added linguistic knowledge but from some simpler regularizing effect we induced on self-attention matrices.
Deep learning has shown unprecedented success in a variety of applications, such as computer vision and medical image analysis. However, there is still potential to improve segmentation in multimodal images by embedding prior knowledge via learning-based shape modeling and registration to learn the modality invariant anatomical structure of organs. For example, in radiotherapy automatic prostate segmentation is essential in prostate cancer diagnosis, therapy, and post-therapy assessment from T2-weighted MR or CT images. In this paper, we present a fully automatic deep generative model-driven multimodal prostate segmentation method using convolutional neural network (DGMNet). The novelty of our method comes with its embedded generative neural network for learning-based shape modeling and its ability to adapt for different imaging modalities via learning-based registration. The proposed method includes a multi-task learning framework that combines a convolutional feature extraction and an embedded regression and classification based shape modeling. This enables the network to predict the deformable shape of an organ. We show that generative neural networkbased shape modeling trained on a reliable contrast imaging modality (such as MRI) can be directly applied to low contrast imaging modality (such as CT) to achieve accurate prostate segmentation. The method was evaluated on MRI and CT datasets acquired from different clinical centers with large variations in contrast and scanning protocols. Experimental results reveal that our method can be used to automatically and accurately segment the prostate gland in different imaging modalities.
We present the results of the Machine Reading for Question Answering (MRQA) 2019 shared task on evaluating the generalization capabilities of reading comprehension systems. In this task, we adapted and unified 18 distinct question answering datasets into the same format. Among them, six datasets were made available for training, six datasets were made available for development, and the final six were hidden for final evaluation. Ten teams submitted systems, which explored various ideas including data sampling, multi-task learning, adversarial training and ensembling. The best system achieved an average F1 score of 72.5 on the 12 held-out datasets, 10.7 absolute points higher than our initial baseline based on BERT.
Post-click conversion rate (CVR) estimation is a critical task in e-commerce recommender systems. This task is deemed quite challenging under the industrial setting with two major issues: 1) selection bias caused by user self-selection, and 2) data sparsity due to the rare click events. A successful conversion typically has the following sequential events: "exposure -> click -> conversion". Conventional CVR estimators are trained in the click space, but the inference is done in the entire exposure space. They fail to account for the causes of the missing data and treat them as missing at random. Hence, their estimations are highly likely to deviate from the real values by large. In addition, the data sparsity issue can also handicap many industrial CVR estimators which usually have large parameter spaces. In this paper, we propose two principled, efficient and highly effective CVR estimators for industrial CVR estimation, namely, Multi-IPW and Multi-DR. The proposed models approach the CVR estimation from a causal perspective and account for the causes of missing not at random. In addition, our methods are based on the multi-task learning framework and mitigate the data sparsity issue. Extensive experiments on industrial-level datasets show that our methods outperform the state-of-the-art CVR models.
Recommender system, as an essential part of modern e-commerce, consists of two fundamental modules, namely Click-Through Rate (CTR) and Conversion Rate (CVR) prediction. While CVR has a direct impact on the purchasing volume, its prediction is well-known challenging due to the Sample Selection Bias (SSB) and Data Sparsity (DS) issues. Although existing methods, typically built on the user sequential behavior path ``impression$\to$click$\to$purchase'', is effective for dealing with SSB issue, they still struggle to address the DS issue due to rare purchase training samples. Observing that users always take several purchase-related actions after clicking, we propose a novel idea of post-click behavior decomposition. Specifically, disjoint purchase-related Deterministic Action (DAction) and Other Action (OAction) are inserted between click and purchase in parallel, forming a novel user sequential behavior graph ``impression$\to$click$\to$D(O)Action$\to$purchase''. Defining model on this graph enables to leverage all the impression samples over the entire space and extra abundant supervised signals from D(O)Action, which will effectively address the SSB and DS issues together. To this end, we devise a novel deep recommendation model named Elaborated Entire Space Supervised Multi-task Model ($ESM^{2}$). According to the conditional probability rule defined on the graph, it employs multi-task learning to predict some decomposed sub-targets in parallel and compose them sequentially to formulate the final CVR. Extensive experiments on both offline and online environments demonstrate the superiority of $ESM^{2}$ over state-of-the-art models. The source code and dataset will be released.
Sporting events are extremely complex and require a multitude of metrics to accurate describe the event. When making multiple predictions, one should make them from a single source to keep consistency across the predictions. We present a multi-task learning method of generating multiple predictions for analysis via a single prediction source. To enable this approach, we utilize a fine-grain representation using fine-grain spatial data using a wide-and-deep learning approach. Additionally, our approach can predict distributions rather than single point values. We highlighted the utility of our approach on the sport of Rugby League and call our prediction engine "Rugby-Bot".
Self-supervised learning, which learns by constructing artificial labels given only the input signals, has recently gained considerable attention for learning representations with unlabeled datasets, i.e., learning without any human-annotated supervision. In this paper, we show that such a technique can be used to significantly improve the model accuracy even under fully-labeled datasets. Our scheme trains the model to learn both original and self-supervised tasks, but is different from conventional multi-task learning frameworks that optimize the summation of their corresponding losses. Our main idea is to learn a single unified task with respect to the joint distribution of the original and self-supervised labels, i.e., we augment original labels via self-supervision of input transformation. This simple, yet effective approach allows to train models easier by relaxing a certain invariant constraint during learning the original and self-supervised tasks simultaneously. It also enables an aggregated inference which combines the predictions from different augmentations to improve the prediction accuracy. Furthermore, we propose a novel knowledge transfer technique, which we refer to as self-distillation, that has the effect of the aggregated inference in a single (faster) inference. We demonstrate the large accuracy improvement and wide applicability of our framework on various fully-supervised settings, e.g., the few-shot and imbalanced classification scenarios.
Affective computing has been largely limited in terms of available data resources. The need to collect and annotate diverse in-the-wild datasets has become apparent with the rise of deep learning models, as the default approach to address any computer vision task. Some in-the-wild databases have been recently proposed. However: i) their size is small, ii) they are not audiovisual, iii) only a small part is manually annotated, iv) they contain a small number of subjects, or v) they are not annotated for all main behavior tasks (valence-arousal estimation, action unit detection and basic expression classification). To address these, we substantially extend the largest available in-the-wild database (Aff-Wild) to study continuous emotions such as valence and arousal. Furthermore, we annotate parts of the database with basic expressions and action units. As a consequence, for the first time, this allows the joint study of all three types of behavior states. We call this database Aff-Wild2. We conduct extensive experiments with CNN and CNN-RNN architectures that use visual and audio modalities; these networks are trained on Aff-Wild2 and their performance is then evaluated on 10 publicly available emotion databases. We show that the networks achieve state-of-the-art performance for the emotion recognition tasks. Additionally, we adapt the ArcFace loss function in the emotion recognition context and use it for training two new networks on Aff-Wild2 and then re-train them in a variety of diverse expression recognition databases. The networks are shown to improve the existing state-of-the-art. The database, emotion recognition models and source code are available at http://ibug.doc.ic.ac.uk/resources/aff-wild2.
The typical multi-task learning methods for spatio-temporal data prediction involve low-rank tensor computation. However, such a method have relatively weak performance when the task number is small, and we cannot integrate it into non-linear models. In this paper, we propose a two-step suboptimal unitary method (SUM) to combine a meta-learning strategy into multi-task models. In the first step, it searches for a global pattern by optimising the general parameters with gradient descents under constraints, which is a geological regularizer to enable model learning with less training data. In the second step, we derive an optimised model on each specific task from the global pattern with only a few local training data. Compared with traditional multi-task learning methods, SUM shows advantages of generalisation ability on distant tasks. It can be applied on any multi-task models with the gradient descent as its optimiser regardless if the prediction function is linear or not. Moreover, we can harness the model to enable traditional prediction model to make coKriging. The experiments on public datasets have suggested that our framework, when combined with current multi-task models, has a conspicuously better prediction result when the task number is small compared to low-rank tensor learning, and our model has a quite satisfying outcome when adjusting the current prediction models for coKriging.
We consider the problem of conversational question answering over a large-scale knowledge base. To handle huge entity vocabulary of a large-scale knowledge base, recent neural semantic parsing based approaches usually decompose the task into several subtasks and then solve them sequentially, which leads to following issues: 1) errors in earlier subtasks will be propagated and negatively affect downstream ones; and 2) each subtask cannot naturally share supervision signals with others. To tackle these issues, we propose an innovative multi-task learning framework where a pointer-equipped semantic parsing model is designed to resolve coreference in conversations, and naturally empower joint learning with a novel type-aware entity detection model. The proposed framework thus enables shared supervisions and alleviates the effect of error propagation. Experiments on a large-scale conversational question answering dataset containing 1.6M question answering pairs over 12.8M entities show that the proposed framework improves overall F1 score from 67% to 79% compared with previous state-of-the-art work.
This paper proposes a novel learning method for multi-task applications. Multi-task neural networks can learn to transfer knowledge across different tasks by using parameter sharing. However, sharing parameters between unrelated tasks can hurt performance. To address this issue, we propose a framework to learn fine-grained patterns of parameter sharing. Assuming that the network is composed of several components across layers, our framework uses learned binary variables to allocate components to tasks in order to encourage more parameter sharing between related tasks, and discourage parameter sharing otherwise. The binary allocation variables are learned jointly with the model parameters by standard back-propagation thanks to the Gumbel-Softmax reparametrization method. When applied to the Omniglot benchmark, the proposed method achieves a 17% relative reduction of the error rate compared to state-of-the-art.
Matrix completion constantly receives tremendous attention from many research fields. It is commonly applied for recommender systems such as movie ratings, computer vision such as image reconstruction or completion, multi-task learning such as collaboratively modeling time-series trends of multiple sensors, and many other applications. Matrix completion techniques are usually computationally exhaustive and/or fail to capture the heterogeneity in the data. For example, images usually contain a heterogeneous set of objects, and thus it is a challenging task to reconstruct images with high levels of missing data. In this paper, we propose the sparse reverse of principal component analysis for matrix completion. The proposed approach maintains smoothness across the matrix, produces accurate estimates of the missing data, converges iteratively, and it is computationally tractable with a controllable upper bound on the number of iterations until convergence. The accuracy of the proposed technique is validated on natural images, movie ratings, and multisensor data. It is also compared with common benchmark methods used for matrix completion.
Detecting manipulated facial images and videos is an increasingly important topic in digital media forensics. As advanced face synthesis and manipulation methods are made available, new types of fake face representations are being created which have raised significant concerns for their use in social media. Hence, it is crucial to detect manipulated face images and localize manipulated regions. Instead of simply using multi-task learning to simultaneously detect manipulated images and predict the manipulated mask (regions), we propose to utilize an attention mechanism to process and improve the feature maps for the classification task. The learned attention maps highlight the informative regions to further improve the binary classification (genuine face v. fake face), and also visualize the manipulated regions. To enable our study of manipulated face detection and localization, we collect a large-scale database that contains numerous types of facial forgeries. With this dataset, we perform a thorough analysis of data-driven fake face detection. We show that the use of an attention mechanism improves facial forgery detection and manipulated region localization.
Federated Learning (FL) is currently the most widely adopted framework for collaborative training of (deep) machine learning models under privacy constraints. Albeit it's popularity, it has been observed that Federated Learning yields suboptimal results if the local clients' data distributions diverge. To address this issue, we present Clustered Federated Learning (CFL), a novel Federated Multi-Task Learning (FMTL) framework, which exploits geometric properties of the FL loss surface, to group the client population into clusters with jointly trainable data distributions. In contrast to existing FMTL approaches, CFL does not require any modifications to the FL communication protocol to be made, is applicable to general non-convex objectives (in particular deep neural networks) and comes with strong mathematical guarantees on the clustering quality. CFL is flexible enough to handle client populations that vary over time and can be implemented in a privacy preserving way. As clustering is only performed after Federated Learning has converged to a stationary point, CFL can be viewed as a post-processing method that will always achieve greater or equal performance than conventional FL by allowing clients to arrive at more specialized models. We verify our theoretical analysis in experiments with deep convolutional and recurrent neural networks on commonly used Federated Learning datasets.
Registration is a fundamental task in medical image analysis which can be applied to several tasks including image segmentation, intra-operative tracking, multi-modal image alignment, and motion analysis. Popular registration tools such as ANTs and NiftyReg optimize an objective function for each pair of images from scratch which is time-consuming for large images with complicated deformation. Facilitated by the rapid progress of deep learning, learning-based approaches such as VoxelMorph have been emerging for image registration. These approaches can achieve competitive performance in a fraction of a second on advanced GPUs. In this work, we construct a neural registration framework, called NeurReg, with a hybrid loss of displacement fields and data similarity, which substantially improves the current state-of-the-art of registrations. Within the framework, we simulate various transformations by a registration simulator which generates fixed image and displacement field ground truth for training. Furthermore, we design three segmentation frameworks based on the proposed registration framework: 1) atlas-based segmentation, 2) joint learning of both segmentation and registration tasks, and 3) multi-task learning with atlas-based segmentation as an intermediate feature. Extensive experimental results validate the effectiveness of the proposed NeurReg framework based on various metrics: the endpoint error (EPE) of the predicted displacement field, mean square error (MSE), normalized local cross-correlation (NLCC), mutual information (MI), Dice coefficient, uncertainty estimation, and the interpretability of the segmentation. The proposed NeurReg improves registration accuracy with fast inference speed, which can greatly accelerate related medical image analysis tasks.
This paper describes the proposed system of the Hitachi team for the Cross-Framework Meaning Representation Parsing (MRP 2019) shared task. In this shared task, the participating systems were asked to predict nodes, edges and their attributes for five frameworks, each with different order of "abstraction" from input tokens. We proposed a unified encoder-to-biaffine network for all five frameworks, which effectively incorporates a shared encoder to extract rich input features, decoder networks to generate anchorless nodes in UCCA and AMR, and biaffine networks to predict edges. Our system was ranked fifth with the macro-averaged MRP F1 score of 0.7604, and outperformed the baseline unified transition-based MRP. Furthermore, post-evaluation experiments showed that we can boost the performance of the proposed system by incorporating multi-task learning, whereas the baseline could not. These imply efficacy of incorporating the biaffine network to the shared architecture for MRP and that learning heterogeneous meaning representations at once can boost the system performance.
Machine Reading Comprehension (MRC) for question answering (QA), which aims to answer a question given the relevant context passages, is an important way to test the ability of intelligence systems to understand human language. Multiple-Choice QA (MCQA) is one of the most difficult tasks in MRC because it often requires more advanced reading comprehension skills such as logical reasoning, summarization, and arithmetic operations, compared to the extractive counterpart where answers are usually spans of text within given passages. Moreover, most existing MCQA datasets are small in size, making the learning task even harder. We introduce MMM, a Multi-stage Multi-task learning framework for Multi-choice reading comprehension. Our method involves two sequential stages: coarse-tuning stage using out-of-domain datasets and multi-task learning stage using a larger in-domain dataset to help model generalize better with limited data. Furthermore, we propose a novel multi-step attention network (MAN) as the top-level classifier for this task. We demonstrate MMM significantly advances the state-of-the-art on four representative MCQA datasets.
Amharic is the official language of the Federal Democratic Republic of Ethiopia. There are lots of historic Amharic and Ethiopic handwritten documents addressing various relevant issues including governance, science, religious, social rules, cultures and art works which are very reach indigenous knowledge. The Amharic language has its own alphabet derived from Ge'ez which is currently the liturgical language in Ethiopia. Handwritten character recognition for non Latin scripts like Amharic is not addressed especially using the advantages of the state of the art techniques. This research work designs for the first time a model for Amharic handwritten character recognition using a convolutional neural network. The dataset was organized from collected sample handwritten documents and data augmentation was applied for machine learning. The model was further enhanced using multi-task learning from the relationships of the characters. Promising results are observed from the later model which can further be applied to word prediction.