In display advertising, predicting the conversion rate, that is, the probability that a user takes a predefined action on an advertiser's website is fundamental in estimating the value of showing a user an advertisement. There are two troublesome difficulties in the conversion rate prediction due to the delayed feedback. First, some positive labels are not correctly observed in training data, because some conversions do not occur right after clicking the ads. Moreover, the delay mechanism is not uniform among instances; some positive feedback is much more frequently observed than the others. It is widely acknowledged that these problems cause a severe bias in the naive empirical average loss function for the conversion rate prediction. To overcome the challenges, we propose two unbiased estimators, one for the conversion rate prediction, and the other for the bias estimation. Subsequently, we propose an interactive learning algorithm named {\em Dual Learning Algorithm for Delayed Feedback (DLA-DF)} where a conversion rate predictor and a bias estimator are learned alternately. The proposed algorithm is the first of its kind to address the two major challenges in a theoretically principal way. Lastly, we conducted a simulation experiment to demonstrate that the proposed method outperforms the existing baselines and validate that the unbiased estimation approach is suitable for the delayed feedback problem.

We study the iteration complexity of stochastic gradient descent (SGD) for minimizing the gradient norm of smooth, possibly nonconvex functions. We provide several results, implying that the classical $\mathcal{O}(\epsilon^{-4})$ upper bound (for making the average gradient norm less than $\epsilon$) cannot be improved upon, unless a combination of additional assumptions is made. Notably, this holds even if we limit ourselves to convex quadratic functions. We also show that for nonconvex functions, the feasibility of minimizing gradients with SGD is surprisingly sensitive to the choice of optimality criteria.

Deep neural networks (DNNs) have been shown to over-fit a dataset when being trained with noisy labels for a long enough time. To overcome this problem, we present a simple and effective method self-ensemble label filtering (SELF) to progressively filter out the wrong labels during training. Our method improves the task performance by gradually allowing supervision only from the potentially non-noisy (clean) labels and stops learning on the filtered noisy labels. For the filtering, we form running averages of predictions over the entire training dataset using the network output at different training epochs. We show that these ensemble estimates yield more accurate identification of inconsistent predictions throughout training than the single estimates of the network at the most recent training epoch. While filtered samples are removed entirely from the supervised training loss, we dynamically leverage them via semi-supervised learning in the unsupervised loss. We demonstrate the positive effect of such an approach on various image classification tasks under both symmetric and asymmetric label noise and at different noise ratios. It substantially outperforms all previous works on noise-aware learning across different datasets and can be applied to a broad set of network architectures. The acquisition of large quantities of a high-quality human annotation is a frequent bottleneck in applying DNNs. There are two cheap but imperfect alternatives to collect annotation at large scale: crowdsourcing from non-experts and web annotations, particularly for image data where the tags and online query keywords are treated as valid labels. Both these alternatives typically introduce noisy (wrong) labels. While Rolnick et al. (2017) empirically demonstrated that DNNs can be surprisingly robust to label noise under certain conditions, Zhang et al. (2017) has shown that DNNs have the capacity to memorize the data and will do so eventually when being confronted with too many noisy labels. Consequently, training DNNs with traditional learning procedures on noisy data strongly deteriorates their ability to generalize - a severe problem.

The data arising in many application domains have an intrinsic network structure. Such network structure is computationally apprealing due to the availability of highly scalable graph algorithms. An important class of graph algorithms is related to optimizing network flows. This paper explores the duality of network flow methods and the recently proposed network Lasso. Network Lasso extends the Lasso method from sparse linear models to clustered graph signals. It turns out that the computational and statistical properties of network Lasso crucially depends on the existence of sufficiently large network flows. Using elementary tools from convex analysis, we offer a precise characterization of the duality between network Lasso and a minimum cost network flow problem. This duality provides a strong link between network Lasso methods and network flow algorithms.

W e derive streamlined mean field variational Bayes algorithms for fitting linear mixed models with crossed random effects. In the most general situation, where the dimensions of the crossed groups are arbitrarily large, streamlining is hindered by lack of sparseness in the underlying least squares system. Because of this fact we also consider a hierarchy of relaxations of the mean field product restriction. The least stringent product restriction delivers a high degree of inferential accuracy . However, this accuracy must be mitigated against its higher storage and computing demands. Faster sparse storage and computing alternatives are also provided, but come with the price of diminished inferential accuracy . This article provides full algorithmic details of three variational inference strategies, presents detailed empirical results on their pros and cons and, thus, guides the users on their choice of variational inference approach depending on the problem size and computing resources. Keywords: Mean field variational Bayes; item response theory; Rasch analysis; scalable statistical methodology; sparse least squares systems.

In this work, we present a novel upper bound of target error to address the problem for unsupervised domain adaptation. Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks. Furthermore, a theory proposed by Ben-David et al. (2010) provides a upper bound for target error when transferring the knowledge, which can be summarized as minimizing the source error and distance between marginal distributions simultaneously. However, common methods based on the theory usually ignore the joint error such that samples from different classes might be mixed together when matching marginal distribution. And in such case, no matter how we minimize the marginal discrepancy, the target error is not bounded due to an increasing joint error. To address this problem, we propose a general upper bound taking joint error into account, such that the undesirable case can be properly penalized. In addition, we utilize constrained hypothesis space to further formalize a tighter bound as well as a novel cross margin discrepancy to measure the dissimilarity between hypotheses which alleviates instability during adversarial learning. Extensive empirical evidence shows that our proposal outperforms related approaches in image classification error rates on standard domain adaptation benchmarks.

In recent years, machine learning has made leaps and bounds enabling applications with high recognition accuracy for speech and images. However, other types of data to which these models can be applied have not yet been explored as thoroughly. In particular, it can be relatively challenging to accurately classify single or multi-model, real-time sensor data. Labelling is an indispensable stage of data pre-processing that can be even more challenging in real-time sensor data collection. Currently, real-time sensor data labelling is an unwieldly process with limited tools available and poor performance characteristics that can lead to the performance of the machine learning models being compromised. In this paper, we introduce new techniques for labelling at the point of collection coupled with a systematic performance comparison of two popular types of Deep Neural Networks running on five custom built edge devices. These state-of-the-art edge devices are designed to enable real-time labelling with various buttons, slide potentiometer and force sensors. This research provides results and insights that can help researchers utilising edge devices for real-time data collection select appropriate labelling techniques. We also identify common bottlenecks in each architecture and provide field tested guidelines to assist developers building adaptive, high performance edge solutions.

The Social Internet of Things (SIoT), integration of Internet of Things and Social networks paradigms, has been introduced to build a network of smart nodes which are capable of establishing social links. In order to deal with misbehavioral service provider nodes, service requestor nodes must evaluate their trustworthiness levels. In this paper, we propose a novel trust management mechanism in the SIoT to predict the most reliable service provider for a service requestor, that leads to reduce the risk of exposing to malicious nodes. We model an SIoT with a flexible bipartite graph (containing two sets of nodes: service providers and requestors), then build the corresponding social network among service requestor nodes, using Hellinger distance. After that, we develop a social trust model, by using nodes' centrality and similarity measures, to extract behavioral trust between the network nodes. Finally, a matrix factorization technique is designed to extract latent features of SIoT nodes to mitigate the data sparsity and cold start problems. We analyze the effect of parameters in the proposed trust prediction mechanism on prediction accuracy. The results indicate that feedbacks from the neighboring nodes of a specific service requestor with high Hellinger similarity in our mechanism outperforms the best existing methods. We also show that utilizing social trust model, which only considers the similarity measure, significantly improves the accuracy of the prediction mechanism. Furthermore, we evaluate the effectiveness of the proposed trust management system through a real-world SIoT application. Our results demonstrate that the proposed mechanism is resilient to different types of network attacks and it can accurately find the proper service provider with high trustworthiness.

This paper explores a novel approach to achieving emergent compositional communication in multi-agent systems. We propose a training regime implementing template transfer, the idea of carrying over learned biases across contexts. In our method, a sender-receiver pair is first trained with disentangled loss functions and then the receiver is transferred to train a new sender with a standard loss. Unlike other methods (e.g. the obverter algorithm), our approach does not require imposing inductive biases on the architecture of the agents. We experimentally show the emergence of compositional communication using topographical similarity, zero-shot generalization and context independence as evaluation metrics. The presented approach is connected to an important line of work in semiotics and developmental psycholinguistics: it supports a conjecture that compositional communication is scaffolded on simpler communication protocols.

Computing expected predictions has many interesting applications in areas such as fairness, handling missing values, and data analysis. Unfortunately, computing expectations of a discriminative model with respect to a probability distribution defined by an arbitrary generative model has been proven to be hard in general. In fact, the task is intractable even for simple models such as logistic regression and a naive Bayes distribution. In this paper, we identify a pair of generative and discriminative models that enables tractable computation of expectations of the latter with respect to the former, as well as moments of any order, in case of regression. Specifically, we consider expressive probabilistic circuits with certain structural constraints that support tractable probabilistic inference. Moreover, we exploit the tractable computation of high-order moments to derive an algorithm to approximate the expectations, for classification scenarios in which exact computations are intractable. We evaluate the effectiveness of our exact and approximate algorithms in handling missing data during prediction time where they prove to be competitive to standard imputation techniques on a variety of datasets. Finally, we illustrate how expected prediction framework can be used to reason about the behaviour of discriminative models.