Guntermcnulty4240

g. feature maps of previous tasks) via attention mechanism. The experiments on variants of MNIST, CIFAR-100 and Sequence of 5-Datasets demonstrate that our methods outperform the state-of-the-art in preventing catastrophic forgetting and fitting new tasks better under the same or less computing resource.AutoML aims at best configuring learning systems automatically. It contains core subtasks of algorithm selection and hyper-parameter tuning. Previous approaches considered searching in the joint hyper-parameter space of all algorithms, which forms a huge but redundant space and causes an inefficient search. We tackle this issue in a \emphcascaded algorithm selection way, which contains an upper-level process of algorithm selection and a lower-level process of hyper-parameter tuning for algorithms. While the lower-level process employs an \emphanytime tuning approach, the upper-level process is naturally formulated as a multi-armed bandit, deciding which algorithm should be allocated one more piece of time for the lower-level tuning. To achieve the goal of finding the best configuration, we propose the \emphExtreme-Region Upper Confidence Bound (ER-UCB) strategy. Unlike UCB bandits that maximize the mean of feedback distribution, ER-UCB maximizes the extreme-region of feedback distribution. We firstly consider stationary distributions and propose the ER-UCB-S algorithm that has O(Klnn) regret upper bound with K arms and n trials. We then extend to non-stationary settings and propose the ER-UCB-N algorithm that has O(Knν) regret upper bound, where [Formula see text]. Finally, empirical studies on synthetic and AutoML tasks verify the effectiveness of ER-UCB-S/N by their outperformance in corresponding settings.We consider the problem of predicting a response Y from a set of covariates X when test- and training distributions differ. Since such differences may have causal explanations, we consider test distributions that emerge from interventions in a structural causal model, and focus on minimizing the worst-case risk. Causal regression models, which regress the response on its direct causes, remain unchanged under arbitrary interventions on the covariates, but they are not always optimal in the above sense. For example, for linear models and bounded interventions, alternative solutions have been shown to be minimax prediction optimal. We introduce the formal framework of distribution generalization that allows us to analyze the above problem in partially observed nonlinear models for both direct interventions on X and interventions that occur indirectly via exogenous variables A. It takes into account that, in practice, minimax solutions need to be identified from data. Our framework allows us to characterize under which class of interventions the causal function is minimax optimal. We prove sufficient conditions for distribution generalization and present corresponding impossibility results. We propose a practical method, NILE, that achieves distribution generalization in a nonlinear IV setting with linear extrapolation. We prove consistency and present empirical results.Noisy labels often occur in vision datasets, especially when they are obtained from crowdsourcing or Web scraping. We propose a new regularization method, which enables learning robust classifiers in presence of noisy data. To achieve this goal, we propose a new adversarial regularization scheme based on the Wasserstein distance. Using this distance allows taking into account specific relations between classes by leveraging the geometric properties of the labels space. Our Wasserstein Adversarial Regularization (WAR) encodes a selective regularization, which promotes smoothness of the classifier between some classes, while preserving sufficient complexity of the decision boundary between others. We first discuss how and why adversarial regularization can be used in the context of noise and then show the effectiveness of our method on five datasets corrupted with noisy labels in both benchmarks and real datasets, WAR outperforms the state-of-the-art competitors.One of the most prominent attributes of Neural Networks (NNs) constitutes their capability of learning to extract robust and descriptive features from high dimensional data, like images. Hence, such an ability renders their exploitation as feature extractors particularly frequent in an abundant of modern reasoning systems. Their application scope mainly includes complex cascade tasks, like multi-modal recognition and deep Reinforcement Learning (RL). However, NNs induce implicit biases that are difficult to avoid or to deal with and are not met in traditional image descriptors. Moreover, the lack of knowledge for describing the intra-layer properties -and thus their general behavior- restricts the further applicability of the extracted features. With the paper at hand, a novel way of visualizing and understanding the vector space before the NNs output layer is presented, aiming to enlighten the deep feature vectors properties under classification tasks. Main attention is paid to the nature of overfitting in the feature space and its adverse effect on further exploitation. We present the findings that can be derived from our models formulation, and we evaluate them on realistic recognition scenarios, proving its prominence by improving the obtained results.With the increasing social demands of disaster response, methods of visual observation for rescue and safety have become increasingly important. However, because of the shortage of datasets for disaster scenarios, there has been little progress in computer vision and robotics in this field. With this in mind, we present the first large-scale synthetic dataset of egocentric viewpoints for disaster scenarios. We simulate pre- and post-disaster cases with drastic changes in appearance, such as buildings on fire and earthquakes. The dataset consists of more than 300K high-resolution stereo image pairs, all annotated with ground-truth data for the semantic label, depth in metric scale, optical flow with sub-pixel precision, and surface normal as well as their corresponding camera poses. To create realistic disaster scenes, we manually augment the effects with 3D models using physically-based graphics tools. We train various state-of-the-art methods to perform computer vision tasks using our dataset, evaluate how well these methods recognize the disaster situations, and produce reliable results of virtual scenes as well as real-world images. We also present a convolutional neural network-based egocentric localization method that is robust to drastic appearance changes, such as the texture changes in a fire, and layout changes from a collapse. To address these key challenges, we propose a new model that learns a shape-based representation by training on stylized images, and incorporate the dominant planes of query images as approximate scene coordinates. We evaluate the proposed method using various scenes including a simulated disaster dataset to demonstrate the effectiveness of our method when confronted with significant changes in scene layout. Experimental results show that our method provides reliable camera pose predictions despite vastly changed conditions.

Presbyopia, an age-related ocular disorder, is characterized by the loss in the accommodative abilities of the human eye. Conventional methods of correcting presbyopia divide the field of view, thereby resulting in significant vision impairment. We demonstrate the design, assembly and evaluation of autofocusing eyeglasses for restoration of accommodation without dividing the field of view.

The adaptive optics eyeglasses comprise of two variable-focus liquid lenses, a time-of-flight range sensor and low-power, dual microprocessor control electronics, housed within an ergonomic frame. Subject-specific accommodation deficiency models were utilized to demonstrate high-fidelity accommodative correction. The abilities of this system to reduce accommodation deficiency, its power consumption, response time, optical performance and MTF were evaluated.

Average corrected accommodation deficiencies for 5 subjects ranged from -0.021 D to 0.016 D. Each accommodation correction calculation was performed in ~67 ms which consumed 4.86 mJ of energy. The optical resolution of the system was 10.5 cycles/degree, and featured a restorative accommodative range of 4.3 D. This system was capable of running for up to 19 hours between charge cycles and weighed ~132 g.

The design, assembly and performance of an autofocusing eyeglasses system to restore accommodation in presbyopes has been demonstrated.

The new autofocusing eyeglasses system presented in this article has the potential to restore pre-presbyopic levels of accommodation in subjects diagnosed with presbyopia.

The new autofocusing eyeglasses system presented in this article has the potential to restore pre-presbyopic levels of accommodation in subjects diagnosed with presbyopia.Mobile genetic elements, elements that can move horizontally between genomes, have profound effects on their host's fitness. The phage-inducible chromosomal island-like element (PLE) is a mobile element that integrates into the chromosome of Vibrio cholerae and parasitizes the bacteriophage ICP1 to move between cells. selleck chemical This parasitism by PLE is such that it abolishes the production of ICP1 progeny and provides a defensive boon to the host cell population. In response to the severe parasitism imposed by PLE, ICP1 has acquired an adaptive CRISPR-Cas system that targets the PLE genome during infection. However, ICP1 isolates that naturally lack CRISPR-Cas are still able to overcome certain PLE variants, and the mechanism of this immunity against PLE has thus far remained unknown. Here, we show that ICP1 isolates that lack CRISPR-Cas encode an endonuclease in the same locus, and that the endonuclease provides ICP1 with immunity to a subset of PLEs. Further analysis shows that this endonuclease is of chimeric origin, incorporating a DNA-binding domain that is highly similar to some PLE replication origin-binding proteins. This similarity allows the endonuclease to bind and cleave PLE origins of replication. The endonuclease appears to exert considerable selective pressure on PLEs and may drive PLE replication module swapping and origin restructuring as mechanisms of escape. This work demonstrates that new genome defense systems can arise through domain shuffling and provides a greater understanding of the evolutionary forces driving genome modularity and temporal succession in mobile elements.Hantaviruses are RNA viruses with known epidemic threat and potential for emergence. Several rodent-borne hantaviruses cause zoonoses accompanied by severe illness and death. However, assessments of zoonotic risk and the development of countermeasures are challenged by our limited knowledge of the molecular mechanisms of hantavirus infection, including the identities of cell entry receptors and their roles in influencing viral host range and virulence. Despite the long-standing presumption that β3/β1-containing integrins are the major hantavirus entry receptors, rigorous genetic loss-of-function evidence supporting their requirement, and that of decay-accelerating factor (DAF), is lacking. Here, we used CRISPR/Cas9 engineering to knockout candidate hantavirus receptors, singly and in combination, in a human endothelial cell line that recapitulates the properties of primary microvascular endothelial cells, the major targets of viral infection in humans. The loss of β3 integrin, β1 integrin, and/or DAF had little or no effect on entry by a large panel of hantaviruses.