The Transformer-based encoder-decoder architecture has recently made significant advances in recognizing handwritten mathematical expressions. However, the transformer model still suffers from the lack of coverage problem, making its expression recognition rate (ExpRate) inferior to its RNN counterpart. Coverage information, which records the alignment information of the past steps, has proven effective in the RNN models. In this paper, we propose CoMER, a model that adopts the coverage information in the transformer decoder. Specifically, we propose a novel Attention Refinement Module (ARM) to refine the attention weights with past alignment information without hurting its parallelism. Furthermore, we take coverage information to the extreme by proposing self-coverage and cross-coverage, which utilize the past alignment information from the current and previous layers. Experiments show that CoMER improves the ExpRate by 0.61%/2.09%/1.59% compared to the current state-of-the-art model, and reaches 59.33%/59.81%/62.97% on the CROHME 2014/2016/2019 test sets.
Conversational recommendation systems (CRS) engage with users by inferring user preferences from dialog history, providing accurate recommendations, and generating appropriate responses. Previous CRSs use knowledge graph (KG) based recommendation modules and integrate KG with language models for response generation. Although KG-based approaches prove effective, two issues remain to be solved. First, KG-based approaches ignore the information in the conversational context but only rely on entity relations and bag of words to recommend items. Second, it requires substantial engineering efforts to maintain KGs that model domain-specific relations, thus leading to less flexibility. In this paper, we propose a simple yet effective architecture comprising a pre-trained language model (PLM) and an item metadata encoder. The encoder learns to map item metadata to embeddings that can reflect the semantic information in the dialog context. The PLM then consumes the semantic-aligned item embeddings together with dialog context to generate high-quality recommendations and responses. Instead of modeling entity relations with KGs, our model reduces engineering complexity by directly converting each item to an embedding. Experimental results on the benchmark dataset ReDial show that our model obtains state-of-the-art results on both recommendation and response generation tasks.
Connectivity and controllability of a complex network are two important issues that guarantee a networked system to function. Robustness of connectivity and controllability guarantees the system to function properly and stably under various malicious attacks. Evaluating network robustness using attack simulations is time consuming, while the convolutional neural network (CNN)-based prediction approach provides a cost-efficient method to approximate the network robustness. In this paper, we investigate the performance of CNN-based approaches for connectivity and controllability robustness prediction, when partial network information is missing, namely the adjacency matrix is incomplete. Extensive experimental studies are carried out. A threshold is explored that if a total amount of more than 7.29\% information is lost, the performance of CNN-based prediction will be significantly degenerated for all cases in the experiments. Two scenarios of missing edge representations are compared, 1) a missing edge is marked `no edge' in the input for prediction, and 2) a missing edge is denoted using a special marker of `unknown'. Experimental results reveal that the first representation is misleading to the CNN-based predictors.
Among various sensors for assisted and autonomous driving systems, automotive radar has been considered as a robust and low-cost solution even in adverse weather or lighting conditions. With the recent development of radar technologies and open-sourced annotated data sets, semantic segmentation with radar signals has become very promising. However, existing methods are either computationally expensive or discard significant amounts of valuable information from raw 3D radar signals by reducing them to 2D planes via averaging. In this work, we introduce ERASE-Net, an Efficient RAdar SEgmentation Network to segment the raw radar signals semantically. The core of our approach is the novel detect-then-segment method for raw radar signals. It first detects the center point of each object, then extracts a compact radar signal representation, and finally performs semantic segmentation. We show that our method can achieve superior performance on radar semantic segmentation task compared to the state-of-the-art (SOTA) technique. Furthermore, our approach requires up to 20x less computational resources. Finally, we show that the proposed ERASE-Net can be compressed by 40% without significant loss in performance, significantly more than the SOTA network, which makes it a more promising candidate for practical automotive applications.
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.
Collaborative table-carrying is a complex task due to the continuous nature of the action and state-spaces, multimodality of strategies, existence of obstacles in the environment, and the need for instantaneous adaptation to other agents. In this work, we present a method for predicting realistic motion plans for cooperative human-robot teams on a table-carrying task. Using a Variational Recurrent Neural Network, VRNN, to model the variation in the trajectory of a human-robot team over time, we are able to capture the distribution over the team's future states while leveraging information from interaction history. The key to our approach is in our model's ability to leverage human demonstration data and generate trajectories that synergize well with humans during test time. We show that the model generates more human-like motion compared to a baseline, centralized sampling-based planner, Rapidly-exploring Random Trees (RRT). Furthermore, we evaluate the VRNN planner with a human partner and show its ability to both generate more human-like paths and achieve higher task success rate than RRT can while planning with a human. Finally, we demonstrate that a LoCoBot using the VRNN planner can complete the task successfully with a human controlling another LoCoBot.
Small target detection is an essential yet challenging task in defense applications, since differentiating low-contrast targets from natural textured and noisy environment remains difficult. To better take into account the contextual information, we propose to explore deep learning approaches based on attention mechanisms. Specifically, we propose a customized version of TransUnet including channel attention, which has shown a significant improvement in performance. Moreover, the lack of annotated data induces weak detection precision, leading to many false alarms. We thus explore a contrario methods in order to select meaningful potential targets detected by a weak deep learning training. -- La d\'etection de petites cibles est une probl\'ematique d\'elicate mais essentielle dans le domaine de la d\'efense, notamment lorsqu'il s'agit de diff\'erencier ces cibles d'un fond bruit\'e ou textur\'e, ou lorsqu'elles sont de faible contraste. Pour mieux prendre en compte les informations contextuelles, nous proposons d'explorer diff\'erentes approches de segmentation par apprentissage profond, dont certaines bas\'ees sur les m\'ecanismes d'attention. Nous proposons \'egalement d'inclure un module d'attention par canal au TransUnet, r\'eseau \`a l'\'etat de l'art, ce qui permet d'am\'eliorer significativement les performances. Par ailleurs, le manque de donn\'ees annot\'ees induit une perte en pr\'ecision lors des d\'etections, conduisant \`a de nombreuses fausses alarmes non pertinentes. Nous explorons donc des m\'ethodes a contrario afin de s\'electionner les cibles les plus significatives d\'etect\'ees par un r\'eseau entra\^in\'e avec peu de donn\'ees.
Constructing large-scaled medical knowledge graphs can significantly boost healthcare applications for medical surveillance, bring much attention from recent research. An essential step in constructing large-scale MKG is extracting information from medical reports. Recently, information extraction techniques have been proposed and show promising performance in biomedical information extraction. However, these methods only consider limited types of entity and relation due to the noisy biomedical text data with complex entity correlations. Thus, they fail to provide enough information for constructing MKGs and restrict the downstream applications. To address this issue, we propose Biomedical Information Extraction, a hybrid neural network to extract relations from biomedical text and unstructured medical reports. Our model utilizes a multi-head attention enhanced graph convolutional network to capture the complex relations and context information while resisting the noise from the data. We evaluate our model on two major biomedical relationship extraction tasks, chemical-disease relation and chemical-protein interaction, and a cross-hospital pan-cancer pathology report corpus. The results show that our method achieves superior performance than baselines. Furthermore, we evaluate the applicability of our method under a transfer learning setting and show that BioIE achieves promising performance in processing medical text from different formats and writing styles.
In view of the propagation environment directly determining the channel fading, the application tasks can also be solved with the aid of the environment information. Inspired by task-oriented semantic communication and machine learning (ML) powered environment-channel mapping methods, this work aims to provide a new view of the environment from the semantic level, which defines the propagation environment semantics (PES) as a limited set of propagation environment semantic symbols (PESS) for diverse application tasks. The PESS is extracted oriented to the tasks with channel properties as a foundation. For method validation, the PES-aided beam prediction (PESaBP) is presented in non-line-of-sight (NLOS). The PESS of environment features and graphs are given for the semantic actions of channel quality evaluation and target scatterer detection of maximum power, which can obtain 0.92 and 0.9 precision, respectively, and save over 87% of time cost.
Transformers have resulted in remarkable achievements in the field of image processing. Inspired by this great success, the application of Transformers to 3D point cloud processing has drawn more and more attention. This paper presents a novel point cloud representational learning network, 3D Point Cloud Transformer with Dual Self-attention (3DPCT) and an encoder-decoder structure. Specifically, 3DPCT has a hierarchical encoder, which contains two local-global dual-attention modules for the classification task (three modules for the segmentation task), with each module consisting of a Local Feature Aggregation (LFA) block and a Global Feature Learning (GFL) block. The GFL block is dual self-attention, with both point-wise and channel-wise self-attention to improve feature extraction. Moreover, in LFA, to better leverage the local information extracted, a novel point-wise self-attention model, named as Point-Patch Self-Attention (PPSA), is designed. The performance is evaluated on both classification and segmentation datasets, containing both synthetic and real-world data. Extensive experiments demonstrate that the proposed method achieved state-of-the-art results on both classification and segmentation tasks.