Accurate forecasting of medical service requirements is an important big data problem that is crucial for resource management in critical times such as natural disasters and pandemics. With the global spread of coronavirus disease 2019 (COVID-19), several concerns have been raised regarding the ability of medical systems to handle sudden changes in the daily routines of healthcare providers. One significant problem is the management of ambulance dispatch and control during a pandemic. To help address this problem, we first analyze ambulance dispatch data records from April 2014 to August 2020 for Nagoya City, Japan. Significant changes were observed in the data during the pandemic, including the state of emergency (SoE) declared across Japan. In this study, we propose a deep learning framework based on recurrent neural networks to estimate the number of emergency ambulance dispatches (EADs) during a SoE. The fusion of data includes environmental factors, the localization data of mobile phone users, and the past history of EADs, thereby providing a general framework for knowledge discovery and better resource management. The results indicate that the proposed blend of training data can be used efficiently in a real-world estimation of EAD requirements during periods of high uncertainties such as pandemics.

Recent works have demonstrated reasonable success of representation learning in hypercomplex space. Specifically, "fully-connected layers with Quaternions" (4D hypercomplex numbers), which replace real-valued matrix multiplications in fully-connected layers with Hamilton products of Quaternions, both enjoy parameter savings with only 1/4 learnable parameters and achieve comparable performance in various applications. However, one key caveat is that hypercomplex space only exists at very few predefined dimensions (4D, 8D, and 16D). This restricts the flexibility of models that leverage hypercomplex multiplications. To this end, we propose parameterizing hypercomplex multiplications, allowing models to learn multiplication rules from data regardless of whether such rules are predefined. As a result, our method not only subsumes the Hamilton product, but also learns to operate on any arbitrary nD hypercomplex space, providing more architectural flexibility using arbitrarily $1/n$ learnable parameters compared with the fully-connected layer counterpart. Experiments of applications to the LSTM and Transformer models on natural language inference, machine translation, text style transfer, and subject verb agreement demonstrate architectural flexibility and effectiveness of the proposed approach.

Data augmentation technique from computer vision has been widely considered as a regularization method to improve data efficiency and generalization performance in vision-based reinforcement learning. We variate the timing of using augmentation, which is, in turn, critical depending on tasks to be solved in training and testing. According to our experiments on Open AI Procgen Benchmark, if the regularization imposed by augmentation is helpful only in testing, it is better to procrastinate the augmentation after training than to use it during training in terms of sample and computation complexity. We note that some of such augmentations can disturb the training process. Conversely, an augmentation providing regularization useful in training needs to be used during the whole training period to fully utilize its benefit in terms of not only generalization but also data efficiency. These phenomena suggest a useful timing control of data augmentation in reinforcement learning.

Spatial memory, or the ability to remember and recall specific locations and objects, is central to autonomous agents' ability to carry out tasks in real environments. However, most existing artificial memory modules are not very adept at storing spatial information. We propose a parameter-free module, Egospheric Spatial Memory (ESM), which encodes the memory in an ego-sphere around the agent, enabling expressive 3D representations. ESM can be trained end-to-end via either imitation or reinforcement learning, and improves both training efficiency and final performance against other memory baselines on both drone and manipulator visuomotor control tasks. The explicit egocentric geometry also enables us to seamlessly combine the learned controller with other non-learned modalities, such as local obstacle avoidance. We further show applications to semantic segmentation on the ScanNet dataset, where ESM naturally combines image-level and map-level inference modalities. Through our broad set of experiments, we show that ESM provides a general computation graph for embodied spatial reasoning, and the module forms a bridge between real-time mapping systems and differentiable memory architectures. Egocentric spatial memory is central to our understanding of spatial reasoning in biology (Klatzky, 1998; Burgess, 2006), where an embodied agent constantly carries with it a local map of its surrounding geometry. Such representations have particular significance for action selection and motor control (Hinman et al., 2019). For robotics and embodied AI, the benefits of a persistent local spatial memory are also clear. Such a system has the potential to run for long periods, and bypass both the memory and runtime complexities of large scale world-centric mapping. Peters et al. (2001) propose an EgoSphere as being a particularly suitable representation for robotics, and more recent works have utilized ego-centric formulations for planar robot mapping (Fankhauser et al., 2014), drone obstacle avoidance (Fragoso et al., 2018) and mono-to-depth (Liu et al., 2019). In parallel with these ego-centric mapping systems, a new paradigm of differentiable memory architectures has arisen, where a memory bank is augmented to a neural network, which can then learn read and write operations (Weston et al., 2014; Graves et al., 2014; Sukhbaatar et al., 2015). When compared to Recurrent Neural Networks (RNNs), the persistent memory circumvents issues of vanishing or exploding gradients, enabling solutions to long-horizon tasks.

This paper comes to address this gap by presenting an empirical analysis of scalability and performance of inferencing Transfomerbased The Transformer architecture revolutionized the field of natural models on CPUs. We identify the key component of the language processing (NLP). Transformers-based models (e.g., BERT) Transformer architecture where the bulk of the computation happens, power many important Web services, such as search, translation, namely, the matrix multiplication (matmul) operations, and question-answering, etc. While enormous research attention is paid propose three optimizations to speed them up. to the training of those models, relatively little efforts are made The first optimization is based on the observation that the performance to improve their inference performance. This paper comes to address of the matmul operation is heavily impacted not only this gap by presenting an empirical analysis of scalability by the shape (dimensions) of the source matrices and the available and performance of inferencing a Transformer-based model on computing resources (the number of worker threads), but also by CPUs.

Clinical practice in intensive care units (ICUs) requires early warnings when a patient's condition is about to deteriorate so that preventive measures can be undertaken. To this end, prediction algorithms have been developed that estimate the risk of mortality in ICUs. In this work, we propose a novel generative deep probabilistic model for real-time risk scoring in ICUs. Specifically, we develop an attentive deep Markov model called AttDMM. To the best of our knowledge, AttDMM is the first ICU prediction model that jointly learns both long-term disease dynamics (via attention) and different disease states in health trajectory (via a latent variable model). Our evaluations were based on an established baseline dataset (MIMIC-III) with 53,423 ICU stays. The results confirm that compared to state-of-the-art baselines, our AttDMM was superior: AttDMM achieved an area under the receiver operating characteristic curve (AUROC) of 0.876, which yielded an improvement over the state-of-the-art method by 2.2%. In addition, the risk score from the AttDMM provided warnings several hours earlier. Thereby, our model shows a path towards identifying patients at risk so that health practitioners can intervene early and save patient lives.

Energy-efficient buildings are one of the top priorities to sustainably address the global energy demands and reduction of CO2 emissions. Advanced control strategies for buildings have been identified as a potential solution with projected energy saving potential of up to 28%. However, the main bottleneck of the model-free methods such as reinforcement learning (RL) is the sampling inefficiency and thus requirement for large datasets, which are costly to obtain or often not available in the engineering practice. On the other hand, model-based methods such as model predictive control (MPC) suffer from large cost associated with the development of the physics-based building thermal dynamics model. We address the challenge of developing cost and data-efficient predictive models of a building's thermal dynamics via physics-constrained deep learning.

Scientists at the University of California, Irvine have developed a new deep-learning framework that predicts gene regulation at the single-cell level. In a study published recently in Science Advances, UCI researchers describe how their deep-learning technique can also be successfully used to observe gene regulation at the cellular level. Until now, that process had been limited to tissue-level analysis. According to co-author Xiaohui Xie, UCI professor of computer science, the framework enables the study of transcription factor binding at the cellular level, which was previously impossible due to the intrinsic noise and sparsity of single-cell data. A transcription factor (TF) is a protein that controls the translation of genetic information from DNA to RNA; TFs regulate genes to ensure they're expressed in proper sequence and at the right time in cells.

Computer-aided detection (CAD) of benign and malignant breast lesions becomes increasingly essential in breast ultrasound (US) imaging. The CAD systems rely on imaging features identified by the medical experts for their performance, whereas deep learning (DL) methods automatically extract features from the data. The challenge of the DL is the insufficiency of breast US images available to train the DL models. Here, we present an ensemble transfer learning model to classify benign and malignant breast tumors using B-mode breast US (B-US) and strain elastography breast US (SE-US) images. This model combines semantic features from AlexNet & ResNet models to classify benign from malignant tumors. We use both B-US and SE-US images to train the model and classify the tumors. We retrospectively gathered 85 patients' data, with 42 benign and 43 malignant cases confirmed with the biopsy. Each patient had multiple B-US and their corresponding SE-US images, and the total dataset contained 261 B-US images and 261 SE-US images. Experimental results show that our ensemble model achieves a sensitivity of 88.89% and specificity of 91.10%. These diagnostic performances of the proposed method are equivalent to or better than manual identification. Thus, our proposed ensemble learning method would facilitate detecting early breast cancer, reliably improving patient care.

DNA sequencing is becoming increasingly commonplace, both in medical and direct-to-consumer settings. To promote discovery, collected genomic data is often de-identified and shared, either in public repositories, such as OpenSNP, or with researchers through access-controlled repositories. However, recent studies have suggested that genomic data can be effectively matched to high-resolution three-dimensional face images, which raises a concern that the increasingly ubiquitous public face images can be linked to shared genomic data, thereby re-identifying individuals in the genomic data. While these investigations illustrate the possibility of such an attack, they assume that those performing the linkage have access to extremely well-curated data. Given that this is unlikely to be the case in practice, it calls into question the pragmatic nature of the attack. As such, we systematically study this re-identification risk from two perspectives: first, we investigate how successful such linkage attacks can be when real face images are used, and second, we consider how we can empower individuals to have better control over the associated re-identification risk. We observe that the true risk of re-identification is likely substantially smaller for most individuals than prior literature suggests. In addition, we demonstrate that the addition of a small amount of carefully crafted noise to images can enable a controlled trade-off between re-identification success and the quality of shared images, with risk typically significantly lowered even with noise that is imperceptible to humans.