As neural networks have begun performing increasingly critical tasks for society, ranging from driving cars to identifying candidates for drug development, the value of their ability to perform uncertainty quantification (UQ) in their predictions has risen commensurately. Permanent dropout, a popular method for neural network UQ, involves injecting stochasticity into the inference phase of the model and creating many predictions for each of the test data. This shifts the computational and energy burden of deep neural networks from the training phase to the inference phase. Recent work has demonstrated near-lossless conversion of classical deep neural networks to their spiking counterparts. We use these results to demonstrate the feasibility of conducting the inference phase with permanent dropout on spiking neural networks, mitigating the technique's computational and energy burden, which is essential for its use at scale or on edge platforms. We demonstrate the proposed approach via the Nengo spiking neural simulator on a combination drug therapy dataset for cancer treatment, where UQ is critical. Our results indicate that the spiking approximation gives a predictive distribution practically indistinguishable from that given by the classical network.

For any financial organization, computing accurate quarterly forecasts for various products is one of the most critical operations. As the granularity at which forecasts are needed increases, traditional statistical time series models may not scale well. We apply deep neural networks in the forecasting domain by experimenting with techniques from Natural Language Processing (Encoder-Decoder LSTMs) and Computer Vision (Dilated CNNs), as well as incorporating transfer learning. A novel contribution of this paper is the application of curriculum learning to neural network models built for time series forecasting. We illustrate the performance of our models using Microsoft's revenue data corresponding to Enterprise, and Small, Medium & Corporate products, spanning approximately 60 regions across the globe for 8 different business segments, and totaling in the order of tens of billions of USD. We compare our models' performance to the ensemble model of traditional statistics and machine learning techniques currently used by Microsoft Finance. With this in-production model as a baseline, our experiments yield an approximately 30% improvement in overall accuracy on test data. We find that our curriculum learning LSTM-based model performs best, showing that it is reasonable to implement our proposed methods without overfitting on medium-sized data.

Feature crossing captures interactions among categorical features and is useful to enhance learning from tabular data in real-world businesses. In this paper, we present AutoCross, an automatic feature crossing tool provided by 4Paradigm to its customers, ranging from banks, hospitals, to Internet corporations. By performing beam search in a tree-structured space, AutoCross enables efficient generation of high-order cross features, which is not yet visited by existing works. Additionally, we propose successive mini-batch gradient descent and multi-granularity discretization to further improve efficiency and effectiveness, while ensuring simplicity so that no machine learning expertise or tedious hyper-parameter tuning is required. Furthermore, the algorithms are designed to reduce the computational, transmitting, and storage costs involved in distributed computing. Experimental results on both benchmark and real-world business datasets demonstrate the effectiveness and efficiency of AutoCross. It is shown that AutoCross can significantly enhance the performance of both linear and deep models.

This paper addresses the challenging problem of retrieval and matching of graph structured objects, and makes two key contributions. First, we demonstrate how Graph Neural Networks (GNN), which have emerged as an effective model for various supervised prediction problems defined on structured data, can be trained to produce embedding of graphs in vector spaces that enables efficient similarity reasoning. Second, we propose a novel Graph Matching Network model that, given a pair of graphs as input, computes a similarity score between them by jointly reasoning on the pair through a new cross-graph attention-based matching mechanism. We demonstrate the effectiveness of our models on different domains including the challenging problem of control-flow-graph based function similarity search that plays an important role in the detection of vulnerabilities in software systems. The experimental analysis demonstrates that our models are not only able to exploit structure in the context of similarity learning but they can also outperform domain-specific baseline systems that have been carefully hand-engineered for these problems.

Artificial neural networks (ANNs) have achieved significant success in tackling classical and modern machine learning problems. As learning problems grow in scale and complexity, and expand into multi-disciplinary territory, a more modular approach for scaling ANNs will be needed. Modular neural networks (MNNs) are neural networks that embody the concepts and principles of modularity. MNNs adopt a large number of different techniques for achieving modularization. Previous surveys of modularization techniques are relatively scarce in their systematic analysis of MNNs, focusing mostly on empirical comparisons and lacking an extensive taxonomical framework. In this review, we aim to establish a solid taxonomy that captures the essential properties and relationships of the different variants of MNNs. Based on an investigation of the different levels at which modularization techniques act, we attempt to provide a universal and systematic framework for theorists studying MNNs, also trying along the way to emphasise the strengths and weaknesses of different modularization approaches in order to highlight good practices for neural network practitioners.

Recently, several methods have been proposed to explain the predictions of recurrent neural networks (RNNs), in particular of LSTMs. The goal of these methods is to understand the network's decisions by assigning to each input variable, e.g., a word, a relevance indicating to which extent it contributed to a particular prediction. In previous works, some of these methods were not yet compared to one another, or were evaluated only qualitatively. We close this gap by systematically and quantitatively comparing these methods in different settings, namely (1) a toy arithmetic task which we use as a sanity check, (2) a five-class sentiment prediction of movie reviews, and besides (3) we explore the usefulness of word relevances to build sentence-level representations. Lastly, using the method that performed best in our experiments, we show how specific linguistic phenomena such as the negation in sentiment analysis reflect in terms of relevance patterns, and how the relevance visualization can help to understand the misclassification of individual samples.

Relational learning is concerned with methods that cannot only leverage the attributes of data points but also their relationships. Diagnosing a patient, for example, not only depends on the patient's vitals and demographic information but also on the same information about their relatives, the information about the hospitals they have visited, and so on. Relational learning, therefore, does not make the assumption of independence between data points but models their dependency explicitly. Graphs are a natural way to represent relational information and there is a large number of machine learning algorithms leveraging graph structure. Graph neural networks (GNNs) (Scarselli et al., 2009) are one such class of algorithms that are able to incorporate sparse and discrete dependency structures between data points. While a graph structure is available in some domains, in others it has to be inferred or constructed. A possible approach is to first create a k-nearest neighbor (kNN) graph based on some measure of similarity between data points. This is a common strategy used by several learning methods such as LLE (Roweis & Saul, 2000) and Isomap (Tenenbaum et al., 2000).

Our insight is that feed forward networks should be able to learn a decision logic that is captured in the activation patterns of its neurons. We propose to extract such decision patterns that can be considered as invariants of the network with respect to a certain output behavior. We present techniques to extract input invariants as convex predicates on the input space, and layer invariants that represent features captured in the hidden layers. We apply the techniques on the networks for the MNIST and ACASXU applications. Our experiments highlight the use of invariants in a variety of applications, such as explainability, providing robustness guarantees, detecting adversaries, simplifying proofs and network distillation.

Despite its success, deep learning still needs large labeled datasets to succeed. Data augmentation has shown much promise in alleviating the need for more labeled data, but it so far has mostly been applied in supervised settings and achieved limited gains. In this work, we propose to apply data augmentation to unlabeled data in a semi-supervised learning setting. Our method, named Unsupervised Data Augmentation or UDA, encourages the model predictions to be consistent between an unlabeled example and an augmented unlabeled example. Unlike previous methods that use random noise such as Gaussian noise or dropout noise, UDA has a small twist in that it makes use of harder and more realistic noise generated by state-of-the-art data augmentation methods. This small twist leads to substantial improvements on six language tasks and three vision tasks even when the labeled set is extremely small. For example, on the IMDb text classification dataset, with only 20 labeled examples, UDA outperforms the state-of-the-art model trained on 25,000 labeled examples. On standard semi-supervised learning benchmarks, CIFAR-10 with 4,000 examples and SVHN with 1,000 examples, UDA outperforms all previous approaches and reduces more than $30\%$ of the error rates of state-of-the-art methods: going from 7.66% to 5.27% and from 3.53% to 2.46% respectively. UDA also works well on datasets that have a lot of labeled data. For example, on ImageNet, with 1.3M extra unlabeled data, UDA improves the top-1/top-5 accuracy from 78.28/94.36% to 79.04/94.45% when compared to AutoAugment.

In this paper, we consider the problem of open information extraction (OIE) for extracting entity and relation level intermediate structures from sentences in open-domain. We focus on four types of valuable intermediate structures (Relation, Attribute, Description, and Concept), and propose a unified knowledge expression form, SAOKE, to express them. We publicly release a data set which contains more than forty thousand sentences and the corresponding facts in the SAOKE format labeled by crowd-sourcing. To our knowledge, this is the largest publicly available human labeled data set for open information extraction tasks. Using this labeled SAOKE data set, we train an end-to-end neural model using the sequenceto-sequence paradigm, called Logician, to transform sentences into facts. For each sentence, different to existing algorithms which generally focus on extracting each single fact without concerning other possible facts, Logician performs a global optimization over all possible involved facts, in which facts not only compete with each other to attract the attention of words, but also cooperate to share words. An experimental study on various types of open domain relation extraction tasks reveals the consistent superiority of Logician to other states-of-the-art algorithms. The experiments verify the reasonableness of SAOKE format, the valuableness of SAOKE data set, the effectiveness of the proposed Logician model, and the feasibility of the methodology to apply end-to-end learning paradigm on supervised data sets for the challenging tasks of open information extraction.