The accelerating development of autonomous driving technology has placed greater demands on obtaining large amounts of high-quality data. Representative, labeled, real world data serves as the fuel for training deep learning networks, critical for improving self-driving perception algorithms. In this paper, we introduce PandaSet, the first dataset produced by a complete, high-precision autonomous vehicle sensor kit with a no-cost commercial license. The dataset was collected using one 360{\deg} mechanical spinning LiDAR, one forward-facing, long-range LiDAR, and 6 cameras. The dataset contains more than 100 scenes, each of which is 8 seconds long, and provides 28 types of labels for object classification and 37 types of labels for semantic segmentation. We provide baselines for LiDAR-only 3D object detection, LiDAR-camera fusion 3D object detection and LiDAR point cloud segmentation. For more details about PandaSet and the development kit, see https://scale.com/open-datasets/pandaset.
Recent years have witnessed the proliferation of Low-power Wide Area Networks (LPWANs) in the unlicensed band for various Internet-of-Things (IoT) applications. Due to the ultra-low transmission power and long transmission duration, LPWAN devices inevitably suffer from high power Cross Technology Interference (CTI), such as interference from Wi-Fi, coexisting in the same spectrum. To alleviate this issue, this paper introduces the Partial Symbol Recovery (PSR) scheme for improving the CTI resilience of LPWAN. We verify our idea on LoRa, a widely adopted LPWAN technique, as a proof of concept. At the PHY layer, although CTI has much higher power, its duration is relatively shorter compared with LoRa symbols, leaving part of a LoRa symbol uncorrupted. Moreover, due to its high redundancy, LoRa chips within a symbol are highly correlated. This opens the possibility of detecting a LoRa symbol with only part of the chips. By examining the unique frequency patterns in LoRa symbols with time-frequency analysis, our design effectively detects the clean LoRa chips that are free of CTI. This enables PSR to only rely on clean LoRa chips for successfully recovering from communication failures. We evaluate our PSR design with real-world testbeds, including SX1280 LoRa chips and USRP B210, under Wi-Fi interference in various scenarios. Extensive experiments demonstrate that our design offers reliable packet recovery performance, successfully boosting the LoRa packet reception ratio from 45.2% to 82.2% with a performance gain of 1.8 times.
Target-oriented opinion words extraction (TOWE) (Fan et al., 2019b) is a new subtask of target-oriented sentiment analysis that aims to extract opinion words for a given aspect in text. Current state-of-the-art methods leverage position embeddings to capture the relative position of a word to the target. However, the performance of these methods depends on the ability to incorporate this information into word representations. In this paper, we explore a variety of text encoders based on pretrained word embeddings or language models that leverage part-of-speech and position embeddings, aiming to examine the actual contribution of each component in TOWE. We also adapt a graph convolutional network (GCN) to enhance word representations by incorporating syntactic information. Our experimental results demonstrate that BiLSTM-based models can effectively encode position information into word representations while using a GCN only achieves marginal gains. Interestingly, our simple methods outperform several state-of-the-art complex neural structures.
Deep learning based approaches have proven promising to model omics data. However, one of the current limitations compared to statistical and traditional machine learning approaches is the lack of explainability, which not only reduces the reliability, but limits the potential for acquiring novel knowledge from unpicking the "black-box" models. Here we present XOmiVAE, a novel interpretable deep learning model for cancer classification using high-dimensional omics data. XOmiVAE is able to obtain contribution values of each gene and latent dimension for a specific prediction, and the correlation between genes and the latent dimensions. It is also revealed that XOmiVAE can explain both the supervised classification and the unsupervised clustering results from the deep learning network. To the best of our knowledge, XOmiVAE is one of the first activated-based deep learning interpretation method to explain novel clusters generated by variational autoencoders. The results generated by XOmiVAE were validated by both the biomedical knowledge and the performance of downstream tasks. XOmiVAE explanations of deep learning based cancer classification and clustering aligned with current domain knowledge including biological annotation and literature, which shows great potential for novel biomedical knowledge discovery from deep learning models. The top XOmiVAE selected genes and dimensions shown significant influence to the performance of cancer classification. Additionally, we offer important steps to consider when interpreting deep learning models for tumour classification. For instance, we demonstrate the importance of choosing background samples that makes biological sense and the limitations of connection weight based methods to explain latent dimensions.
Digital pathology slide is easy to store and manage, convenient to browse and transmit. However, because of the high-resolution scan for example 40 times magnification(40X) during the digitization, the file size of each whole slide image exceeds 1Gigabyte, which eventually leads to huge storage capacity and very slow network transmission. We design a strategy to scan slides with low resolution (5X) and a super-resolution method is proposed to restore the image details when in diagnosis. The method is based on a multi-scale generative adversarial network, which sequentially generate three high-resolution images such as 10X, 20X and 40X. The perceived loss, generator loss of the generated images and real images are compared on three image resolutions, and a discriminator is used to evaluate the difference of highest-resolution generated image and real image. A dataset consisting of 100,000 pathological images from 10 types of human tissues is performed for training and testing the network. The generated images have high peak-signal-to-noise-ratio (PSNR) and structural-similarity-index (SSIM). The PSNR of 10X to 40X image are 24.16, 22.27 and 20.44, and the SSIM are 0.845, 0.680 and 0.512, which are better than other super-resolution networks such as DBPN, ESPCN, RDN, EDSR and MDSR. Moreover, visual inspections show that the generated high-resolution images by our network have enough details for diagnosis, good color reproduction and close to real images, while other five networks are severely blurred, local deformation or miss important details. Moreover, no significant differences can be found on pathological diagnosis based on the generated and real images. The proposed multi-scale network can generate good high-resolution pathological images, and will provide a low-cost storage (about 15MB/image on 5X), faster image sharing method for digital pathology.
Text retrieval using learned dense representations has recently emerged as a promising alternative to "traditional" text retrieval using sparse bag-of-words representations. One recent work that has garnered much attention is the dense passage retriever (DPR) technique proposed by Karpukhin et al. (2020) for end-to-end open-domain question answering. We present a replication study of this work, starting with model checkpoints provided by the authors, but otherwise from an independent implementation in our group's Pyserini IR toolkit and PyGaggle neural text ranking library. Although our experimental results largely verify the claims of the original paper, we arrived at two important additional findings that contribute to a better understanding of DPR: First, it appears that the original authors under-report the effectiveness of the BM25 baseline and hence also dense--sparse hybrid retrieval results. Second, by incorporating evidence from the retriever and an improved answer span scoring technique, we are able to improve end-to-end question answering effectiveness using exactly the same models as in the original work.
Pansharpening is a fundamental issue in remote sensing field. This paper proposes a side information partially guided convolutional sparse coding (SCSC) model for pansharpening. The key idea is to split the low resolution multispectral image into a panchromatic image related feature map and a panchromatic image irrelated feature map, where the former one is regularized by the side information from panchromatic images. With the principle of algorithm unrolling techniques, the proposed model is generalized as a deep neural network, called as SCSC pansharpening neural network (SCSC-PNN). Compared with 13 classic and state-of-the-art methods on three satellites, the numerical experiments show that SCSC-PNN is superior to others. The codes are available at https://github.com/xsxjtu/SCSC-PNN.
Pan-sharpening is an important technique for remote sensing imaging systems to obtain high resolution multispectral images. Recently, deep learning has become the most popular tool for pan-sharpening. This paper develops a model-based deep pan-sharpening approach. Specifically, two optimization problems regularized by the deep prior are formulated, and they are separately responsible for the generative models for panchromatic images and low resolution multispectral images. Then, the two problems are solved by a gradient projection algorithm, and the iterative steps are generalized into two network blocks. By alternatively stacking the two blocks, a novel network, called gradient projection based pan-sharpening neural network, is constructed. The experimental results on different kinds of satellite datasets demonstrate that the new network outperforms state-of-the-art methods both visually and quantitatively. The codes are available at https://github.com/xsxjtu/GPPNN.
High-dimensional omics data contains intrinsic biomedical information that is crucial for personalised medicine. Nevertheless, it is challenging to capture them from the genome-wide data due to the large number of molecular features and small number of available samples, which is also called "the curse of dimensionality" in machine learning. To tackle this problem and pave the way for machine learning aided precision medicine, we proposed a unified multi-task deep learning framework called OmiEmbed to capture a holistic and relatively precise profile of phenotype from high-dimensional omics data. The deep embedding module of OmiEmbed learnt an omics embedding that mapped multiple omics data types into a latent space with lower dimensionality. Based on the new representation of multi-omics data, different downstream networks of OmiEmbed were trained together with the multi-task strategy to predict the comprehensive phenotype profile of each sample. We trained the model on two publicly available omics datasets to evaluate the performance of OmiEmbed. The OmiEmbed model achieved promising results for multiple downstream tasks including dimensionality reduction, tumour type classification, multi-omics integration, demographic and clinical feature reconstruction, and survival prediction. Instead of training and applying different downstream networks separately, the multi-task strategy combined them together and conducted multiple tasks simultaneously and efficiently. The model achieved better performance with the multi-task strategy comparing to training them individually. OmiEmbed is a powerful tool to accurately capture comprehensive phenotypic information from high-dimensional omics data and has a great potential to facilitate more accurate and personalised clinical decision making.
In spite of much recent research in the area, it is still unclear whether subject-area question-answering data is useful for machine reading comprehension (MRC) tasks. In this paper, we investigate this question. We collect a large-scale multi-subject multiple-choice question-answering dataset, ExamQA, and use incomplete and noisy snippets returned by a web search engine as the relevant context for each question-answering instance to convert it into a weakly-labeled MRC instance. We then propose a self-teaching paradigm to better use the generated weakly-labeled MRC instances to improve a target MRC task. Experimental results show that we can obtain an improvement of 5.1% in accuracy on a multiple-choice MRC dataset, C^3, demonstrating the effectiveness of our framework and the usefulness of large-scale subject-area question-answering data for machine reading comprehension.