Climate models play a crucial role in understanding the effect of environmental and man-made changes on climate to help mitigate climate risks and inform governmental decisions. Large global climate models such as the Community Earth System Model (CESM), developed by the National Center for Atmospheric Research, are very complex with millions of lines of code describing interactions of the atmosphere, land, oceans, and ice, among other components. As development of the CESM is constantly ongoing, simulation outputs need to be continuously controlled for quality. To be able to distinguish a "climate-changing" modification of the code base from a true climate-changing physical process or intervention, there needs to be a principled way of assessing statistical reproducibility that can handle both spatial and temporal high-dimensional simulation outputs. Our proposed work uses probabilistic classifiers like tree-based algorithms and deep neural networks to perform a statistically rigorous goodness-of-fit test of high-dimensional spatio-temporal data.

The aim of this paper is to propose a suitable method for constructing prediction intervals for the output of neural network models. To do this, we adapt the extremely randomized trees method originally developed for random forests to construct ensembles of neural networks. The extra-randomness introduced in the ensemble reduces the variance of the predictions and yields gains in out-of-sample accuracy. An extensive Monte Carlo simulation exercise shows the good performance of this novel method for constructing prediction intervals in terms of coverage probability and mean square prediction error. This approach is superior to state-of-the-art methods extant in the literature such as the widely used MC dropout and bootstrap procedures. The out-of-sample accuracy of the novel algorithm is further evaluated using experimental settings already adopted in the literature.

Information theory is an outstanding framework to measure uncertainty, dependence and relevance in data and systems. It has several desirable properties for real world applications: it naturally deals with multivariate data, it can handle heterogeneous data types, and the measures can be interpreted in physical units. However, it has not been adopted by a wider audience because obtaining information from multidimensional data is a challenging problem due to the curse of dimensionality. Here we propose an indirect way of computing information based on a multivariate Gaussianization transform. Our proposal mitigates the difficulty of multivariate density estimation by reducing it to a composition of tractable (marginal) operations and simple linear transformations, which can be interpreted as a particular deep neural network. We introduce specific Gaussianization-based methodologies to estimate total correlation, entropy, mutual information and Kullback-Leibler divergence. We compare them to recent estimators showing the accuracy on synthetic data generated from different multivariate distributions. We made the tools and datasets publicly available to provide a test-bed to analyze future methodologies. Results show that our proposal is superior to previous estimators particularly in high-dimensional scenarios; and that it leads to interesting insights in neuroscience, geoscience, computer vision, and machine learning.

The COVID-19 pandemic has profound global consequences on health, economic, social, political, and almost every major aspect of human life. Therefore, it is of great importance to model COVID-19 and other pandemics in terms of the broader social contexts in which they take place. We present the architecture of AICov, which provides an integrative deep learning framework for COVID-19 forecasting with population covariates, some of which may serve as putative risk factors. We have integrated multiple different strategies into AICov, including the ability to use deep learning strategies based on LSTM and even modeling. To demonstrate our approach, we have conducted a pilot that integrates population covariates from multiple sources. Thus, AICov not only includes data on COVID-19 cases and deaths but, more importantly, the population's socioeconomic, health and behavioral risk factors at a local level. The compiled data are fed into AICov, and thus we obtain improved prediction by integration of the data to our model as compared to one that only uses case and death data.

Twitter has become an important communication channel in times of emergency. The ubiquitousness of smartphones enables people to announce an emergency they're observing in real-time. Because of this, more agencies are interested in programatically monitoring Twitter (disaster relief organizations and news agencies) and therefore recognizing the informativeness of a tweet can help filter noise from large volumes of data. In this paper, we present our submission for WNUT-2020 Task 2: Identification of informative COVID-19 English Tweets. Our most successful model is an ensemble of transformers including RoBERTa, XLNet, and BERTweet trained in a semi-supervised experimental setting. The proposed system achieves a F1 score of 0.9011 on the test set (ranking 7th on the leaderboard), and shows significant gains in performance compared to a baseline system using fasttext embeddings.

Biological neural networks are equipped with an inherent capability to continuously adapt through online learning. This aspect remains in stark contrast to learning with error backpropagation through time (BPTT) applied to recurrent neural networks (RNNs), or recently to biologically-inspired spiking neural networks (SNNs). BPTT involves offline computation of the gradients due to the requirement to unroll the network through time. Online learning has recently regained the attention of the research community, focusing either on approaches that approximate BPTT or on biologically-plausible schemes applied to SNNs. Here we present an alternative perspective that is based on a clear separation of spatial and temporal gradient components. Combined with insights from biology, we derive from first principles a novel online learning algorithm for deep SNNs, called online spatio-temporal learning (OSTL). For shallow networks, OSTL is gradient-equivalent to BPTT enabling for the first time online training of SNNs with BPTT-equivalent gradients. In addition, the proposed formulation unveils a class of SNN architectures trainable online at low time complexity. Moreover, we extend OSTL to a generic form, applicable to a wide range of network architectures, including networks comprising long short-term memory (LSTM) and gated recurrent units (GRU). We demonstrate the operation of our algorithm on various tasks from language modelling to speech recognition and obtain results on par with the BPTT baselines. The proposed algorithm provides a framework for developing succinct and efficient online training approaches for SNNs and in general deep RNNs.

Although much progress has been made towards robust deep learning, a significant gap in robustness remains between real-world perturbations and more narrowly defined sets typically studied in adversarial defenses. In this paper, we aim to bridge this gap by learning perturbation sets from data, in order to characterize real-world effects for robust training and evaluation. Specifically, we use a conditional generator that defines the perturbation set over a constrained region of the latent space. We formulate desirable properties that measure the quality of a learned perturbation set, and theoretically prove that a conditional variational autoencoder naturally satisfies these criteria. Using this framework, our approach can generate a variety of perturbations at different complexities and scales, ranging from baseline spatial transformations, through common image corruptions, to lighting variations. We measure the quality of our learned perturbation sets both quantitatively and qualitatively, finding that our models are capable of producing a diverse set of meaningful perturbations beyond the limited data seen during training. Finally, we leverage our learned perturbation sets to train models which are empirically and certifiably robust to adversarial image corruptions and adversarial lighting variations, while improving generalization on non-adversarial data. All code and configuration files for reproducing the experiments as well as pretrained model weights can be found at https://github.com/locuslab/perturbation_learning.

There are three main reasons We present LCE, a Local Cascade Ensemble for that justify the use of ensembles over single classifiers [Dietterich, traditional (tabular) multivariate data classification, 2000]: statistical (reduce the risk of choosing the and its extension LCEM for Multivariate Time Series wrong classifier by averaging when the amount of training (MTS) classification. LCE is a new hybrid ensemble data available is too small compared to the size of the hypothesis method that combines an explicit boostingbagging space), computational (local search from many different approach to handle the bias-variance tradeoff starting points may provide a better approximation to faced by machine learning models and an implicit the true unknown function than any of the individual classifier), divide-and-conquer approach to individualize and representational (expansion of the space of representable classifier errors on different parts of the training functions).

Attribution methods assess the contribution of inputs to the model prediction. One way to do so is erasure: a subset of inputs is considered irrelevant if it can be removed without affecting the prediction. Though conceptually simple, erasure's objective is intractable and approximate search remains expensive with modern deep NLP models. Erasure is also susceptible to the hindsight bias: the fact that an input can be dropped does not mean that the model `knows' it can be dropped. The resulting pruning is over-aggressive and does not reflect how the model arrives at the prediction. To deal with these challenges, we introduce Differentiable Masking. DiffMask learns to mask-out subsets of the input while maintaining differentiability. The decision to include or disregard an input token is made with a simple model based on intermediate hidden layers of the analyzed model. First, this makes the approach efficient because we predict rather than search. Second, as with probing classifiers, this reveals what the network `knows' at the corresponding layers. This lets us not only plot attribution heatmaps but also analyze how decisions are formed across network layers. We use DiffMask to study BERT models on sentiment classification and question answering.

Neural models have yielded state-of-the-art results in deciphering spoken language understanding (SLU) problems; however, these models require a significant amount of domain-specific labeled examples for training, which is prohibitively expensive. While pre-trained language models like BERT have been shown to capture a massive amount of knowledge by learning from unlabeled corpora and solve SLU using fewer labeled examples for adaption, the encoding of knowledge is implicit and agnostic to downstream tasks. Such encoding results in model inefficiencies in parameter usage: an entirely new model is required for every domain. To address these challenges, we introduce a novel SLU framework, comprising a conversational language modeling (CLM) pre-training task and a light encoder architecture. The CLM pre-training enables networks to capture the representation of the language in conversation style with the presence of ASR errors. The light encoder architecture separates the shared pre-trained networks from the mappings of generally encoded knowledge to specific domains of SLU, allowing for the domain adaptation to be performed solely at the light encoder and thus increasing efficiency. With the framework, we match the performance of state-of-the-art SLU results on Alexa internal datasets and on two public ones (ATIS, SNIPS), adding only 4.4% parameters per task.