Multi-Task Learning (MTL) has achieved great success in various fields, however, how to balance different tasks to avoid negative effects is still a key problem. To achieve the task balancing, there exist many works to balance task losses or gradients. In this paper, we unify eight representative task balancing methods from the perspective of loss weighting and provide a consistent experimental comparison. Moreover, we surprisingly find that training a MTL model with random weights sampled from a distribution can achieve comparable performance over state-of-the-art baselines. Based on this finding, we propose a simple yet effective weighting strategy called Random Loss Weighting (RLW), which can be implemented in only one additional line of code over existing works. Theoretically, we analyze the convergence of RLW and reveal that RLW has a higher probability to escape local minima than existing models with fixed task weights, resulting in a better generalization ability. Empirically, we extensively evaluate the proposed RLW method on six image datasets and four multilingual tasks from the XTREME benchmark to show the effectiveness of the proposed RLW strategy when compared with state-of-the-art strategies.
In recent years, Multi-Task Learning (MTL) attracts much attention due to its good performance in many applications. However, many existing MTL models cannot guarantee that its performance is no worse than its single-task counterpart on each task. Though this phenomenon has been empirically observed by some works, little work aims to handle the resulting problem, which is formally defined as negative sharing in this paper. To achieve safe multi-task learning where no \textit{negative sharing} occurs, we propose a Safe Multi-Task Learning (SMTL) model, which consists of a public encoder shared by all the tasks, private encoders, gates, and private decoders. Specifically, each task has a private encoder, a gate, and a private decoder, where the gate is to learn how to combine the private encoder and public encoder for the downstream private decoder. To reduce the storage cost during the inference stage, a lite version of SMTL is proposed to allow the gate to choose either the public encoder or the corresponding private encoder. Moreover, we propose a variant of SMTL to place all the gates after decoders of all the tasks. Experiments on several benchmark datasets demonstrate the effectiveness of the proposed methods.
Self-supervised training has shown promising gains in pretraining models and facilitating the downstream finetuning for speech recognition, like multilingual ASR. Most existing methods adopt a 2-stage scheme where the self-supervised loss is optimized in the first pretraining stage, and the standard supervised finetuning resumes in the second stage. In this paper, we propose an end-to-end (E2E) Joint Unsupervised and Supervised Training (JUST) method to combine the supervised RNN-T loss and the self-supervised contrastive and masked language modeling (MLM) losses. We validate its performance on the public dataset Multilingual LibriSpeech (MLS), which includes 8 languages and is extremely imbalanced. On MLS, we explore (1) JUST trained from scratch, and (2) JUST finetuned from a pretrained checkpoint. Experiments show that JUST can consistently outperform other existing state-of-the-art methods, and beat the monolingual baseline by a significant margin, demonstrating JUST's capability of handling low-resource languages in multilingual ASR. Our average WER of all languages outperforms average monolingual baseline by 33.3%, and the state-of-the-art 2-stage XLSR by 32%. On low-resource languages like Polish, our WER is less than half of the monolingual baseline and even beats the supervised transfer learning method which uses external supervision.
The improvement of pose estimation accuracy is currently the fundamental problem in mobile robots. This study aims to improve the use of observations to enhance accuracy. The selection of feature points affects the accuracy of pose estimation, leading to the question of how the contribution of observation influences the system. Accordingly, the contribution of information to the pose estimation process is analyzed. Moreover, the uncertainty model, sensitivity model, and contribution theory are formulated, providing a method for calculating the contribution of every residual term. The proposed selection method has been theoretically proven capable of achieving a global statistical optimum. The proposed method is tested on artificial data simulations and compared with the KITTI benchmark. The experiments revealed superior results in contrast to ALOAM and MLOAM. The proposed algorithm is implemented in LiDAR odometry and LiDAR Inertial odometry both indoors and outdoors using diverse LiDAR sensors with different scan modes, demonstrating its effectiveness in improving pose estimation accuracy. A new configuration of two laser scan sensors is subsequently inferred. The configuration is valid for three-dimensional pose localization in a prior map and yields results at the centimeter level.
Real-time location inference of social media users is the fundamental of some spatial applications such as localized search and event detection. While tweet text is the most commonly used feature in location estimation, most of the prior works suffer from either the noise or the sparsity of textual features. In this paper, we aim to tackle these two problems. We use topic modeling as a building block to characterize the geographic topic variation and lexical variation so that "one-hot" encoding vectors will no longer be directly used. We also incorporate other features which can be extracted through the Twitter streaming API to overcome the noise problem. Experimental results show that our RATE algorithm outperforms several benchmark methods, both in the precision of region classification and the mean distance error of latitude and longitude regression.
We study the problem of weakly supervised text classification, which aims to classify text documents into a set of pre-defined categories with category surface names only and without any annotated training document provided. Most existing approaches leverage textual information in each document. However, in many domains, documents are accompanied by various types of metadata (e.g., authors, venue, and year of a research paper). These metadata and their combinations may serve as strong category indicators in addition to textual contents. In this paper, we explore the potential of using metadata to help weakly supervised text classification. To be specific, we model the relationships between documents and metadata via a heterogeneous information network. To effectively capture higher-order structures in the network, we use motifs to describe metadata combinations. We propose a novel framework, named MotifClass, which (1) selects category-indicative motif instances, (2) retrieves and generates pseudo-labeled training samples based on category names and indicative motif instances, and (3) trains a text classifier using the pseudo training data. Extensive experiments on real-world datasets demonstrate the superior performance of MotifClass to existing weakly supervised text classification approaches. Further analysis shows the benefit of considering higher-order metadata information in our framework.
Advancing lithium-ion batteries (LIBs) in both design and usage is key to promoting electrification in the coming decades to mitigate human-caused climate change. Inadequate understanding of LIB degradation is an important bottleneck that limits battery durability and safety. Here, we propose hybrid physics-based and data-driven modeling for online diagnosis and prognosis of battery degradation. Compared to existing battery modeling efforts, we aim to build a model with physics as its backbone and statistical learning techniques as enhancements. Such a hybrid model has better generalizability and interpretability together with a well-calibrated uncertainty associated with its prediction, rendering it more valuable and relevant to safety-critical applications under realistic usage scenarios.
Unsupervised pre-training is now the predominant approach for both text and speech understanding. Self-attention models pre-trained on large amounts of unannotated data have been hugely successful when fine-tuned on downstream tasks from a variety of domains and languages. This paper takes the universality of unsupervised language pre-training one step further, by unifying speech and text pre-training within a single model. We build a single encoder with the BERT objective on unlabeled text together with the w2v-BERT objective on unlabeled speech. To further align our model representations across modalities, we leverage alignment losses, specifically Translation Language Modeling (TLM) and Speech Text Matching (STM) that make use of supervised speech-text recognition data. We demonstrate that incorporating both speech and text data during pre-training can significantly improve downstream quality on CoVoST~2 speech translation, by around 1 BLEU compared to single-modality pre-trained models, while retaining close to SotA performance on LibriSpeech and SpeechStew ASR tasks. On four GLUE tasks and text-normalization, we observe evidence of capacity limitations and interference between the two modalities, leading to degraded performance compared to an equivalent text-only model, while still being competitive with BERT. Through extensive empirical analysis we also demonstrate the importance of the choice of objective function for speech pre-training, and the beneficial effect of adding additional supervised signals on the quality of the learned representations.
Most research on domain adaptation has focused on the purely unsupervised setting, where no labeled examples in the target domain are available. However, in many real-world scenarios, a small amount of labeled target data is available and can be used to improve adaptation. We address this semi-supervised setting and propose to use dynamic feature alignment to address both inter- and intra-domain discrepancy. Unlike previous approaches, which attempt to align source and target features within a mini-batch, we propose to align the target features to a set of dynamically updated class prototypes, which we use both for minimizing divergence and pseudo-labeling. By updating based on class prototypes, we avoid problems that arise in previous approaches due to class imbalances. Our approach, which doesn't require extensive tuning or adversarial training, significantly improves the state of the art for semi-supervised domain adaptation. We provide a quantitative evaluation on two standard datasets, DomainNet and Office-Home, and performance analysis.
Motivated by the success of T5 (Text-To-Text Transfer Transformer) in pre-training natural language processing models, we propose a unified-modal SpeechT5 framework that explores the encoder-decoder pre-training for self-supervised speech/text representation learning. The SpeechT5 framework consists of a shared encoder-decoder network and six modal-specific (speech/text) pre/post-nets. After preprocessing the speech/text input through the pre-nets, the shared encoder-decoder network models the sequence to sequence transformation, and then the post-nets generate the output in the speech/text modality based on the decoder output. Particularly, SpeechT5 can pre-train on a large scale of unlabeled speech and text data to improve the capability of the speech and textual modeling. To align the textual and speech information into a unified semantic space, we propose a cross-modal vector quantization method with random mixing-up to bridge speech and text. Extensive evaluations on a wide variety of spoken language processing tasks, including voice conversion, automatic speech recognition, text to speech, and speaker identification, show the superiority of the proposed SpeechT5 framework.