教学课程资源库 更多>>

关键词： Computer science   Health sciences   Biostatistics   Medical diagnosis   Datasets   Experiments   Narratives   Neural networks   Classification   Age groups   Annotations  

摘要： The digitization of health records has provided a wealth of data that can be used for machine learning tasks such as automated diagnosis of illness or cause of death. However, crucial contextual information such as time and duration have often been omitted from such models. Although the words and phrases describing time can be modeled with semantic representations such as word embeddings, these representations often fail to capture the temporal value of such time phrases. The goal of this work is to present models for identifying symptoms and time information in the text of different types of health documents (while explicitly modeling temporal value), determining the chronological order of the extracted symptoms, and classifying the documents according to diagnosis or cause-of-death. Since the style and content of health text documents can be widely variable, these models are evaluated on three different types of health records: clinical notes (formal written health records), verbal autopsies (informal written records of deaths that occurred outside of health facilities), and transcripts of clinical conversations (verbal dialogues between medical staff, patients, and caregivers). We conduct end-to-end disease classification using neural network models trained on free-text narratives. The system includes supervised learning models for event and time extraction, listwise temporal ordering of events, and classification. We find that classification models that include event timeline representations perform better than methods that use only the raw text of the narrative. Furthermore, we present several models and metrics for listwise temporal ordering, including the ability to model simultaneous events. These models can be applied to different types of medical narratives, and can improve the performance of automated classifiers which can be used in real-world contexts to improve cause-of-death coding efficiency and clinical decision support.

关键词： Computer science   Artificial intelligence  

摘要： Artificial intelligence is a topic that is fast-moving and ever-changing. A shift is taking place with the availability of complete web services that can be invoked at the click of a button. The power of the home office has been transformed; a technology revolution is taking place. The enterprise information systems (EIS) specialists need to pivot to this new transformation and encompass the vast knowledge required to bring enterprise artificial intelligence systems online. This design science research explores the strategies an EIS specialist can use to invoke a cross-convolutional neural network system. This study includes the Amazon Web Service (AWS), the Ubuntu operating system, PyTorch, and Java. The research aims to aid those new to artificial intelligence and image recognition. An exploration is given of the existing Java-based imaging applications and Java-based image recognition libraries. The key focus of the study is to contribute to the advancement of the field of butterfly image recognition from an EIS perspective. Recommendations within this study detail the steps involved in handling dependencies of libraries and operating systems that are not immediately current. Java mappings and commented source code are also included and detailed from the perspective of an EIS specialist.

关键词： Robotics   Computer science   Artificial intelligence  

摘要： Despite fifty years of research on prosthetic robot hands, this technology is yet to be fully acknowledged by amputees as well as manufacturers, due to a lack of robustness and intuitive control. Our goal is to enhance the control robustness and intuitiveness through integrating the context, i.e. knowledge of environment and task, with human bio-signals, e.g. hand trajectory. Although this solution has already been studied in the literature, no unified framework is proposed for multi-modal information fusion. This research is aimed at introducing a novel framework for human-in-the-loop prosthetic robot hand control. We propose our solution in two parts, (1) grasp inference and (2) end-to-end actuator control. For (1), we propose a model-based and a model-free grasp inference framework. Our model-based method is based on the particle filters method. With knowledge of context hard-coded into the solution per se, our particle-filter-based framework can incorporate any input signal using a proper weight function. In our model-free method, we use the hidden Markov model to learn the grasp-inference task directly from human hand transport trajectories. For (2), we propose a reinforcement learning (RL) framework which learns to control the robot actuator directly from the context with less information hard-coded into the solution. We leverage imitation learning (IL) besides RL to overcome challenging exploration in the problem. To provide invariance to the human hand transport trajectories, we first provide a solution based on synthesized trajectories based on a human motion model. Later, we adopt an over-sampling technique for real human hand transport trajectories, to serve as a means of data augmentation. This research provides a step forward in more rigorous frameworks for multi-modal information fusion for prosthetic robot hand control and grasp inference through model- /data-based methods.

关键词： Star & galaxy formation   Algorithms   Astrophysics   Computer science   Physics  

摘要： The nature of turbulence in molecular clouds is one of the driving factors that influence star formation efficiency. It is speculated that the high star formation efficiency observed in spiral-arm clouds is linked to the prevalence of compressive (curl-free) turbulent modes, while the shear-driven solenoidal (divergence-free) modes appear to be the main cause of the low star formation efficiency that characterises clouds in the Central Molecular Zone (CMZ). Similarly, the analysis of the Orion B molecular cloud confirmed that the dominant solenoidal turbulence is compatible with its low star formation rate. However, turbulent modes vary locally and at different scales within the cloud, and turbulent motions surrounding the main star-forming regions display a strongly compressive nature. This evidence points to inter-and intra-cloud fluctuations of the solenoidal modes being an agent for the variability of star formation efficiency and cloud collision being a facilitator of stars’ formation through the production of highly compressive gas flows. This thesis presents a quantitative estimation of the relative fractions of momentum density in the solenoidal and compressible modes of turbulence in the plane molecular clouds found in the 13CO/C18O (J = 3 → 2) Heterodyne Inner Milky Way Plane Survey (CHIMPS). This calculation is achieved through a statistical method that allows us to reconstruct the 3-dimensional distribution of the density momentum from its line-of-sight projected counterparts (zeroth, first, and second velocity moments) provided by the observations, producing an estimate of the power contained in the solenoidal and compressive turbulent modes within each cloud. The project investigates how different fractions of compressive and solenoidal modes in CHIMPS clouds probe the variation of the star formation efficiency across clouds with varying environments. A negative correlation between the solenoidal fraction and star formation efficiency is found. This fe

关键词： Software packages   Anisotropy   Fault lines   Computer science   Geophysics  

摘要： The ability to accurately and efficiently track material properties in a Lagrangian sense during geodynamical flows, as well as evaluate how they evolve through both space and time, is of vital importance to our understanding of the structure, dynamics and evolution of Earth’s mantle and lithosphere. An approach for achieving this, widely advocated by the geodynamics community, is the so-called particle-in cell technique. With this scheme, material properties are tracked by a large number of particles that are advected with the flow field. These properties can represent a wide range of parameters (e.g. material composition, melt %, material strain and strain history) and the information they carry can be accessed during a simulation to feed back into the flow equations (e.g. composition controlling material density), as well as generate diagnostic fields for analysis (e.g. total melt volume, the generation of lattice-preferred orientation and the generation of synthetic anisotropy predictions). In Chapter 2, we develop a particle-in-cell scheme within an adaptive, unstructured, anisotropic mesh computational modelling framework. Regions of geodynamic interest often vary throughout a simulation, and the combination of the particle-in-cell scheme with a state of the art adaptive mesh algorithm enables both the mesh resolution and particle density to capture areas of interest with high resolution (or density), while reducing resolution (or density) in areas of little geodynamic interest. This ensures that a high level of accuracy is maintained throughout the computational domain, while also greatly improving numerical efficiency. To our knowledge, this is the first example of a particle-in-cell scheme being incorporated within a fully adaptive, unstructured, anisotropic framework in the geodynamics community. The development of the particle-in-cell scheme within this framework saw several inherent challenges, including the interpolation of particle positions from globa

关键词： Biology   Biological oceanography   Computer science   Entomology   Biophysics  

关键词： Transportation   Urban planning   Computer science  

摘要： With the advent of new mobility services and technologies, the complexity of understanding the mobility patterns has been gradually intensified. The availability of large datasets, in conjunction with the transportation revolution, has been increased and incurs high computing costs. These two critical challenges require us to methodologically handle complex transportation problems with numerical performance: fast, high-precision solutions, and reliable structure under different impact factors. That is, it is imperative to introduce a new type of modeling strategy, advancing the conventional transportation planning models. In order to do this, we leverage the backbone of the underlying algorithm behind machine learning (ML): computational graph (CG) and automatic differentiation (AD). CG is a directed acyclic graph (DAG) where each vertex represents a mathematical operation, and each edge represents data transfer. AD is an efficient algorithm to analytically compute gradients of necessary functionality. Embedding the two key algorithms into the planning models, specifically parametric-based econometric models and network optimization models, we theoretically and practically develop different types of modeling structures and reformulate mathematical formulations on basis of the graph-oriented representation. Three closely related analytical and computational frameworks are presented in this dissertation, based on a common modeling methodology of CG abstraction. First, a two-stage interpretable machine learning framework developed by a linear regression model, coupled with a neural network layered by long short-term memory (LSTM) shows the capability of capturing statistical characteristics with enhanced predictability in the context of day-to-day streaming datasets. Second, AD-based computation in estimating for discrete choice models proves more efficiency of handling complex modeling structure than the standard optimization solver relying on numerical gradients, out

关键词： Ocean currents   Wind   Basins   Regions   Maps   Decomposition   Ocean circulation   Energy   Algorithms   Time series   Climate   Computer science   Geomorphology   Geology  

摘要： Over recent decades, changes in the climate system have fundamentally modified properties of the ocean. These adjustments include alterations of the sea level, the sea surface temperature, and the ocean circulation from regional to global scales. The ocean surface responds to changes in the climate system through the exchange of heat, carbon, and momentum. Numerous studies have examined the response of sea level and ocean heat content to climate change; however, it remains unknown how the ocean surface currents have adjusted to the climate system over the past decades. Ocean currents have a wide range of spatial and temporal scales, from submesoscales (100m to 10 km), to mesoscale (10 to 100 km), to large-scale overturning. Dynamically, mesoscale flows are crucial in the transport and mixing of tracers such as heat, salt, and nutrients, but they also constitute one of the major reservoirs of the time-varying kinetic energy, namely eddy kinetic energy (EKE). Mesoscale flows consist of three distinct processes: coherent eddies, jets, and waves. The temporal evolution of the ocean currents and the mesoscale field is yet to be quantified. In this thesis, we use satellite observations and a coherent eddy tracking algorithm (Track: Eddy) to investigate the temporal evolution of the surface mesoscale, coherent eddies, and jets over the last 27 years. We find a multi-decadal readjustment of the eddy and mesoscale field to changes in the climate system from two independent satellite products. Eddy kinetic energy and sea surface temperature gradients show a significant strengthening of the eddy field at rates of 1.2%-1.8% per decade. Furthermore, regions dominated by mesoscale processes such as the Antarctic Circumpolar Current (ACC) and ocean boundary currents have undergone a greater change over the observational record at rates of 5% and 2.5% per decade, respectively. This global and regional readjustment of the surface currents has crucial implications in the exchange of

关键词： Fractals   Near Eastern studies   Computer science   Mechanics   Islamic studies  

摘要： A large part of science involves building and investigating models. One key feature of model-based science is that one thing is studied as a means of learning about some rather different thing. How scientists make inferences from a model to the world, then, is a topic of great interest to philosophers of science. An increasing number of models are specified with very complex computer programs. In this thesis, I examine the epistemological issues that arise when scientists use these computer simulation models to learn about the world or to think through their ideas. I argue that the explosion of computational power over the last several decades has revolutionised model-based science, but that restraint and caution must be exercised in the face of this power. To make my arguments, I focus on two kinds of computer simulation modelling: climate modelling and, in particular, high-fidelity climate models; and agent-based models, which are used to represent populations of interacting agents often in an ecological or social context. Both kinds involve complex model structures and are representative of the beneficial capacities of computer simulation. However, both face epistemic costs that follow from using highly complex model structures. As models increase in size and complexity, it becomes far harder for modellers to understand their models and why they behave the way they do. The value of models is further obscured by their proliferation, and a proliferation of programming languages in which they can be described. If modellers struggle to grasp their models, they can struggle to make good inferences with them. While the climate modelling community has developed much of the infrastructure required to mitigate these epistemic costs, the less mature field of agent-based modelling is still struggling to implement such community standards and infrastructure. I conclude that modellers cannot take full advantage of the representational capacities of computer simulations unless