This paper presents a novel approach for multi-task learning of language understanding (LU) and dialogue state tracking (DST) in task-oriented dialogue systems. Multi-task training enables the sharing of the neural network layers responsible for encoding the user utterance for both LU and DST and improves performance while reducing the number of network parameters. In our proposed framework, DST operates on a set of candidate values for each slot that has been mentioned so far. These candidate sets are generated using LU slot annotations for the current user utterance, dialogue acts corresponding to the preceding system utterance and the dialogue state estimated for the previous turn, enabling DST to handle slots with a large or unbounded set of possible values and deal with slot values not seen during training. Furthermore, to bridge the gap between training and inference, we investigate the use of scheduled sampling on LU output for the current user utterance as well as the DST output for the preceding turn.
One of the key challenges of performing label prediction over a data stream concerns with the emergence of instances belonging to unobserved class labels over time. Previously, this problem has been addressed by detecting such instances and using them for appropriate classifier adaptation. The fundamental aspect of a novel-class detection strategy relies on the ability of comparison among observed instances to discriminate them into known and unknown classes. Therefore, studies in the past have proposed various metrics suitable for comparison over the observed feature space. Unfortunately, these similarity measures fail to reliably identify distinct regions in observed feature spaces useful for class discrimination and novel-class detection, especially in streams containing high-dimensional data instances such as images and texts. In this paper, we address this key challenge by proposing a semi-supervised multi-task learning framework called \sysname{} which aims to intrinsically search for a latent space suitable for detecting labels of instances from both known and unknown classes. We empirically measure the performance of \sysname{} over multiple real-world image and text datasets and demonstrate its superiority by comparing its performance with existing semi-supervised methods.
The key to success in machine learning (ML) is the use of effective data representations. Traditionally, data representations were hand-crafted. Recently it has been demonstrated that, given sufficient data, deep neural networks can learn effective implicit representations from simple input representations. However, for most scientific problems, the use of deep learning is not appropriate as the amount of available data is limited, and/or the output models must be explainable. Nevertheless, many scientific problems do have significant amounts of data available on related tasks, which makes them amenable to multi-task learning, i.e. learning many related problems simultaneously. Here we propose a novel and general representation learning approach for multi-task learning that works successfully with small amounts of data. The fundamental new idea is to transform an input intrinsic data representation (i.e., handcrafted features), to an extrinsic representation based on what a pre-trained set of models predict about the examples. This transformation has the dual advantages of producing significantly more accurate predictions, and providing explainable models. To demonstrate the utility of this transformative learning approach, we have applied it to three real-world scientific problems: drug-design (quantitative structure activity relationship learning), predicting human gene expression (across different tissue types and drug treatments), and meta-learning for machine learning (predicting which machine learning methods work best for a given problem). In all three problems, transformative machine learning significantly outperforms the best intrinsic representation.
We examine two fundamental tasks associated with graph representation learning: link prediction and node classification. We present a new autoencoder architecture capable of learning a joint representation of local graph structure and available node features for the simultaneous multi-task learning of unsupervised link prediction and semi-supervised node classification. Our simple, yet effective and versatile model is efficiently trained end-to-end in a single stage, whereas previous related deep graph embedding methods require multiple training steps that are difficult to optimize. We provide an empirical evaluation of our model on five benchmark relational, graph-structured datasets and demonstrate significant improvement over three strong baselines for graph representation learning. Reference code and data are available at https://github.com/vuptran/graph-representation-learning
The training of many existing end-to-end steering angle prediction models heavily relies on steering angles as the supervisory signal. Without learning from much richer contexts, these methods are susceptible to the presence of sharp road curves, challenging traffic conditions, strong shadows, and severe lighting changes. In this paper, we considerably improve the accuracy and robustness of predictions through heterogeneous auxiliary networks feature mimicking, a new and effective training method that provides us with much richer contextual signals apart from steering direction. Specifically, we train our steering angle predictive model by distilling multi-layer knowledge from multiple heterogeneous auxiliary networks that perform related but different tasks, e.g., image segmentation or optical flow estimation. As opposed to multi-task learning, our method does not require expensive annotations of related tasks on the target set. This is made possible by applying contemporary off-the-shelf networks on the target set and mimicking their features in different layers after transformation. The auxiliary networks are discarded after training without affecting the runtime efficiency of our model. Our approach achieves a new state-of-the-art on Udacity and Comma.ai, outperforming the previous best by a large margin of 12.8% and 52.1%, respectively. Encouraging results are also shown on Berkeley Deep Drive (BDD) dataset.
We frame unsupervised machine translation (MT) in the context of multi-task learning (MTL), combining insights from both directions. We leverage off-the-shelf neural MT architectures to train unsupervised MT models with no parallel data and show that such models can achieve reasonably good performance, competitive with models purpose-built for unsupervised MT. Finally, we propose improvements that allow us to apply our models to English-Turkish, a truly low-resource language pair.
We study approaches to improve fine-grained short answer Question Answering models by integrating coarse-grained data annotated for paragraph-level relevance and show that coarsely annotated data can bring significant performance gains. Experiments demonstrate that the standard multi-task learning approach of sharing representations is not the most effective way to leverage coarse-grained annotations. Instead, we can explicitly model the latent fine-grained short answer variables and optimize the marginal log-likelihood directly or use a newly proposed \emph{posterior distillation} learning objective. Since these latent-variable methods have explicit access to the relationship between the fine and coarse tasks, they result in significantly larger improvements from coarse supervision.
End-to-end Speech Translation (ST) models have many potential advantages when compared to the cascade of Automatic Speech Recognition (ASR) and text Machine Translation (MT) models, including lowered inference latency and the avoidance of error compounding. However, the quality of end-to-end ST is often limited by a paucity of training data, since it is difficult to collect large parallel corpora of speech and translated transcript pairs. Previous studies have proposed the use of pre-trained components and multi-task learning in order to benefit from weakly supervised training data, such as speech-to-transcript or text-to-foreign-text pairs. In this paper, we demonstrate that using pre-trained MT or text-to-speech (TTS) synthesis models to convert weakly supervised data into speech-to-translation pairs for ST training can be more effective than multi-task learning. Furthermore, we demonstrate that a high quality end-to-end ST model can be trained using only weakly supervised datasets, and that synthetic data sourced from unlabeled monolingual text or speech can be used to improve performance. Finally, we discuss methods for avoiding overfitting to synthetic speech with a quantitative ablation study.
Humans have consciousness as the ability to perceive events and objects: a mental model of the world developed from the most impoverished of visual stimuli, enabling humans to make rapid decisions and take actions. Although spatial and temporal aspects of different scenes are generally diverse, the underlying physics among environments still work the same way, thus learning an abstract description of shared physical dynamics helps human to understand the world. In this paper, we explore building this mental world with neural network models through multi-task learning, namely the meta-world model. We show through extensive experiments that our proposed meta-world models successfully capture the common dynamics over the compact representations of visually different environments from Atari Games. We also demonstrate that agents equipped with our meta-world model possess the ability of visual self-recognition, i.e., recognize themselves from the reflected mirrored environment derived from the classic mirror self-recognition test (MSR).
In this work, we propose to model the interaction between visual and textual features for multi-modal neural machine translation (MMT) through a latent variable model. This latent variable can be seen as a multi-modal stochastic embedding of an image and its description in a foreign language. It is used in a target-language decoder and also to predict image features. Importantly, our model formulation utilises visual and textual inputs during training but does not require that images be available at test time. We show that our latent variable MMT formulation improves considerably over strong baselines, including a multi-task learning approach (Elliott and K\'ad\'ar, 2017) and a conditional variational auto-encoder approach (Toyama et al., 2016). Finally, we show improvements due to (i) predicting image features in addition to only conditioning on them, (ii) imposing a constraint on the minimum amount of information encoded in the latent variable, and (iii) by training on additional target-language image descriptions (i.e. synthetic data).
Current Neural Machine Translation (NMT) employs a language-specific encoder to represent the source sentence and adopts a language-specific decoder to generate target translation. This language-dependent design leads to large-scale network parameters and makes the duality of the parallel data underutilized. To address the problem, we propose in this paper a language-independent representor to replace the encoder and decoder by using weight sharing. This shared representor can not only reduce large portion of network parameters, but also facilitate us to fully explore the language duality by jointly training source-to-target, target-to-source, left-to-right and right-to-left translations within a multi-task learning framework. Experiments show that our proposed framework can obtain significant improvements over conventional NMT models on resource-rich and low-resource translation tasks with only a quarter of parameters.
The explosive growth of digital images in video surveillance and social media has led to the significant need for efficient search of persons of interest in law enforcement and forensic applications. Despite tremendous progress in primary biometric traits (e.g., face and fingerprint) based person identification, a single biometric trait alone cannot meet the desired recognition accuracy in forensic scenarios. Tattoos, as one of the important soft biometric traits, have been found to be valuable for assisting in person identification. However, tattoo search in a large collection of unconstrained images remains a difficult problem, and existing tattoo search methods mainly focus on matching cropped tattoos, which is different from real application scenarios. To close the gap, we propose an efficient tattoo search approach that is able to learn tattoo detection and compact representation jointly in a single convolutional neural network (CNN) via multi-task learning. While the features in the backbone network are shared by both tattoo detection and compact representation learning, individual latent layers of each sub-network optimize the shared features toward the detection and feature learning tasks, respectively. We resolve the small batch size issue inside the joint tattoo detection and compact representation learning network via random image stitch and preceding feature buffering. We evaluate the proposed tattoo search system using multiple public-domain tattoo benchmarks, and a gallery set with about 300K distracter tattoo images compiled from these datasets and images from the Internet. In addition, we also introduce a tattoo sketch dataset containing 300 tattoos for sketch-based tattoo search. Experimental results show that the proposed approach has superior performance in tattoo detection and tattoo search at scale compared to several state-of-the-art tattoo retrieval algorithms.
Segmentation of the left atrium (LA) is crucial for assessing its anatomy in both pre-operative atrial fibrillation (AF) ablation planning and post-operative follow-up studies. In this paper, we present a fully automated framework for left atrial segmentation in gadolinium-enhanced magnetic resonance images (GE-MRI) based on deep learning. We propose a fully convolutional neural network and explore the benefits of multi-task learning for performing both atrial segmentation and pre/post ablation classification. Our results show that, by sharing features between related tasks, the network can gain additional anatomical information and achieve more accurate atrial segmentation, leading to a mean Dice score of 0.901 on a test set of 20 3D MRI images. Code of our proposed algorithm is available at https://github.com/cherise215/atria_segmentation_2018/.
This paper investigates the vision-based autonomous driving with deep learning and reinforcement learning methods. Different from the end-to-end learning method, our method breaks the vision-based lateral control system down into a perception module and a control module. The perception module which is based on a multi-task learning neural network first takes a driver-view image as its input and predicts the track features. The control module which is based on reinforcement learning then makes a control decision based on these features. In order to improve the data efficiency, we propose visual TORCS (VTORCS), a deep reinforcement learning environment which is based on the open racing car simulator (TORCS). By means of the provided functions, one can train an agent with the input of an image or various physical sensor measurement, or evaluate the perception algorithm on this simulator. The trained reinforcement learning controller outperforms the linear quadratic regulator (LQR) controller and model predictive control (MPC) controller on different tracks. The experiments demonstrate that the perception module shows promising performance and the controller is capable of controlling the vehicle drive well along the track center with visual input.
We propose a methodology for estimating human behaviors in psychotherapy sessions using mutli-label and multi-task learning paradigms. We discuss the problem of behavioral coding in which data of human interactions is the annotated with labels to describe relevant human behaviors of interest. We describe two related, yet distinct, corpora consisting of therapist client interactions in psychotherapy sessions. We experimentally compare the proposed learning approaches for estimating behaviors of interest in these datasets. Specifically, we compare single and multiple label learning approaches, single and multiple task learning approaches, and evaluate the performance of these approaches when incorporating turn context. We demonstrate the prediction performance gains which can be achieved by using the proposed paradigms and discuss the insights these models provide into these complex interactions.
We introduce a novel method that enables parameter-efficient transfer and multi-task learning with deep neural networks. The basic approach is to learn a model patch - a small set of parameters - that will specialize to each task, instead of fine-tuning the last layer or the entire network. For instance, we show that learning a set of scales and biases is sufficient to convert a pretrained network to perform well on qualitatively different problems (e.g. converting a Single Shot MultiBox Detection (SSD) model into a 1000-class image classification model while reusing 98% of parameters of the SSD feature extractor). Similarly, we show that re-learning existing low-parameter layers (such as depth-wise convolutions) while keeping the rest of the network frozen also improves transfer-learning accuracy significantly. Our approach allows both simultaneous (multi-task) as well as sequential transfer learning. In several multi-task learning problems, despite using much fewer parameters than traditional logits-only fine-tuning, we match single-task performance.
Cardiac magnetic resonance (CMR) is used extensively in the diagnosis and management of cardiovascular disease. Deep learning methods have proven to deliver segmentation results comparable to human experts in CMR imaging, but there have been no convincing results for the problem of end-to-end segmentation and diagnosis from CMR. This is in part due to a lack of sufficiently large datasets required to train robust diagnosis models. In this paper, we propose a learning method to train diagnosis models, where our approach is designed to work with relatively small datasets. In particular, the optimisation loss is based on multi-task learning that jointly trains for the tasks of segmentation and diagnosis classification. We hypothesize that segmentation has a regularizing effect on the learning of features relevant for diagnosis. Using the 100 training and 50 testing samples available from the Automated Cardiac Diagnosis Challenge (ACDC) dataset, which has a balanced distribution of 5 cardiac diagnoses, we observe a reduction of the classification error from 32% to 22%, and a faster convergence compared to a baseline without segmentation. To the best of our knowledge, this is the best diagnosis results from CMR using an end-to-end diagnosis and segmentation learning method.
In this paper, we describe a general framework: Parameters Read-Write Networks (PRaWNs) to systematically analyze current neural models for multi-task learning, in which we find that existing models expect to disentangle features into different spaces while features learned in practice are still entangled in shared space, leaving potential hazards for other training or unseen tasks. We propose to alleviate this problem by incorporating an inductive bias into the process of multi-task learning, that each task can keep informed of not only the knowledge stored in other tasks but the way how other tasks maintain their knowledge. In practice, we achieve above inductive bias by allowing different tasks to communicate by passing both hidden variables and gradients explicitly. Experimentally, we evaluate proposed methods on three groups of tasks and two types of settings (\textsc{in-task} and \textsc{out-of-task}). Quantitative and qualitative results show their effectiveness.
Precision medicine aims for personalized prognosis and therapeutics by utilizing recent genome-scale high-throughput profiling techniques, including next-generation sequencing (NGS). However, translating NGS data faces several challenges. First, NGS count data are often overdispersed, requiring appropriate modeling. Second, compared to the number of involved molecules and system complexity, the number of available samples for studying complex disease, such as cancer, is often limited, especially considering disease heterogeneity. The key question is whether we may integrate available data from all different sources or domains to achieve reproducible disease prognosis based on NGS count data. In this paper, we develop a Bayesian Multi-Domain Learning (BMDL) model that derives domain-dependent latent representations of overdispersed count data based on hierarchical negative binomial factorization for accurate cancer subtyping even if the number of samples for a specific cancer type is small. Experimental results from both our simulated and NGS datasets from The Cancer Genome Atlas (TCGA) demonstrate the promising potential of BMDL for effective multi-domain learning without "negative transfer" effects often seen in existing multi-task learning and transfer learning methods.
Estimating multiple attributes from a single facial image gives comprehensive descriptions on the high level semantics of the face. It is naturally regarded as a multi-task supervised learning problem with a single deep CNN, in which lower layers are shared, and higher ones are task-dependent with the multi-branch structure. Within the traditional deep multi-task learning (DMTL) framework, this paper intends to fully exploit the correlations among different attributes by constructing a graph. The node in graph represents the feature vector from a particular branch for a given attribute, and the edge can be defined by either the prior knowledge or the similarity between two nodes in the embedding with a fully data-driven manner. We analyze that the attention mechanism actually takes effect in the latter case, and utilize the Graph Attention Layer (GAL) for exploring on the most relevant attribute feature and refining the task-dependant feature by considering other attributes. Experiments show that by mining the correlations among attributes, our method can improve the recognition accuracy on CelebA and LFWA dataset. And it also achieves competitive performance.