Various classes of Graph Neural Networks (GNN) have been proposed and shown to be successful in a wide range of applications with graph structured data. In this paper, we propose a theoretical framework able to compare the expressive power of these GNN architectures. The current universality theorems only apply to intractable classes of GNNs. Here, we prove the first approximation guarantees for practical GNNs, paving the way for a better understanding of their generalization. Our theoretical results are proved for invariant GNNs computing a graph embedding (permutation of the nodes of the input graph does not affect the output) and equivariant GNNs computing an embedding of the nodes (permutation of the input permutes the output). We show that Folklore Graph Neural Networks (FGNN), which are tensor based GNNs augmented with matrix multiplication are the most expressive architectures proposed so far for a given tensor order. We illustrate our results on the Quadratic Assignment Problem (a NP-Hard combinatorial problem) by showing that FGNNs are able to learn how to solve the problem, leading to much better average performances than existing algorithms (based on spectral, SDP or other GNNs architectures). On a practical side, we also implement masked tensors to handle batches of graphs of varying sizes.

Along with the proliferation of Artificial Intelligence (AI) and Internet of Things (IoT) techniques, various kinds of adversarial attacks are increasingly emerging to fool Deep Neural Networks (DNNs) used by Industrial IoT (IIoT) applications. Due to biased training data or vulnerable underlying models, imperceptible modifications on inputs made by adversarial attacks may result in devastating consequences. Although existing methods are promising in defending such malicious attacks, most of them can only deal with limited existing attack types, which makes the deployment of large-scale IIoT devices a great challenge. To address this problem, we present an effective federated defense approach named FDA3 that can aggregate defense knowledge against adversarial examples from different sources. Inspired by federated learning, our proposed cloud-based architecture enables the sharing of defense capabilities against different attacks among IIoT devices. Comprehensive experimental results show that the generated DNNs by our approach can not only resist more malicious attacks than existing attack-specific adversarial training methods, but also can prevent IIoT applications from new attacks.

Sparsity has become popular in machine learning, because it can save computational resources, facilitate interpretations, and prevent overfitting. In this paper, we discuss sparsity in the framework of neural networks. In particular, we formulate a new notion of sparsity that concerns the networks' layers and, therefore, aligns particularly well with the current trend toward deep networks. We call this notion layer sparsity. We then introduce corresponding regularization and refitting schemes that can complement standard deep-learning pipelines to generate more compact and accurate networks.

This has made it easier to step outside the rigid model and approximation families designed based on computational tractability, and to switch to flexible distributions parameterized by neural networks, including normalizing flows relying on invertible transformation of simple base distributions [21]. Models with discrete latent variables, however, remain problematic due to non-differentiability of the sampling operation that prevents efficient optimization of expectations over the approximation. Hence in practice, we still largely resort to model-specific algorithms that are tedious to extend already for minor variants of the model, analytic marginalization of the discrete latent variables (e.g., mixture models and LDA in Stan [7]), continuous relaxations like the concrete distribution [19, 13], or (semi-)implicit approximations that do not support probability evaluation [34, 30]. Normalizing flows [25, 15, 21], suitable for learning flexible posterior approximations for continuous variables, have recently been generalized also for discrete categorical [31] and ordinal [12] variables. However, the discrete variants have only been applied in generative modeling of discrete observations. Even though discrete normalizing flows for categorical distributions (DNF) retain the property of differentiable Monte Carlo estimates, they are not very suitable for variational approximation due to their limited expressive power. As we will show later, a DNF can only move probability mass around and hence relies extremely strongly on use of base distributions that are already expressive (and of the same dimension as the final distribution). For generative modeling this can be satisfied, for example by using recurrent neural networks as base distributions in the case of language modeling [31], but for modeling latent variables there are no easy ways of learning strong base distributions. We improve the expressive power of DNFs by constructing mixtures of categorical discrete normalizing flows (MDNF).

Deep generative models (e.g. GANs and VAEs) have been developed quite extensively in recent years. Lately, there has been an increased interest in the inversion of such a model, i.e. given a (possibly corrupted) signal, we wish to recover the latent vector that generated it. Building upon sparse representation theory, we define conditions that are applicable to any inversion algorithm (gradient descent, deep encoder, etc.), under which such generative models are invertible with a unique solution. Importantly, the proposed analysis is applicable to any trained model, and does not depend on Gaussian i.i.d. weights. Furthermore, we introduce two layer-wise inversion pursuit algorithms for trained generative networks of arbitrary depth, and accompany these with recovery guarantees. Finally, we validate our theoretical results numerically and show that our method outperforms gradient descent when inverting such generators, both for clean and corrupted signals.

Learning graph generative models is a challenging task for deep learning and has wide applicability to a range of domains like chemistry, biology and social science. However current deep neural methods suffer from limited scalability: for a graph with $n$ nodes and $m$ edges, existing deep neural methods require $\Omega(n^2)$ complexity by building up the adjacency matrix. On the other hand, many real world graphs are actually sparse in the sense that $m\ll n^2$. Based on this, we develop a novel autoregressive model, named BiGG, that utilizes this sparsity to avoid generating the full adjacency matrix, and importantly reduces the graph generation time complexity to $O((n + m)\log n)$. Furthermore, during training this autoregressive model can be parallelized with $O(\log n)$ synchronization stages, which makes it much more efficient than other autoregressive models that require $\Omega(n)$. Experiments on several benchmarks show that the proposed approach not only scales to orders of magnitude larger graphs than previously possible with deep autoregressive graph generative models, but also yields better graph generation quality.

Rich user behavior data has been proven to be of great value for click-through rate prediction tasks, especially in industrial applications such as recommender systems and online advertising. Both industry and academy have paid much attention to this topic and propose different approaches to modeling with long sequential user behavior data. Among them, memory network based model MIMN proposed by Alibaba, achieves SOTA with the co-design of both learning algorithm and serving system. MIMN is the first industrial solution that can model sequential user behavior data with length scaling up to 1000. However, MIMN fails to precisely capture user interests given a specific candidate item when the length of user behavior sequence increases further, say, by 10 times or more. This challenge exists widely in previously proposed approaches. In this paper, we tackle this problem by designing a new modeling paradigm, which we name as Search-based Interest Model (SIM). SIM extracts user interests with two cascaded search units: (i) General Search Unit acts as a general search from the raw and arbitrary long sequential behavior data, with query information from candidate item, and gets a Sub user Behavior Sequence which is relevant to candidate item; (ii) Exact Search Unit models the precise relationship between candidate item and SBS. This cascaded search paradigm enables SIM with a better ability to model lifelong sequential behavior data in both scalability and accuracy. Apart from the learning algorithm, we also introduce our hands-on experience on how to implement SIM in large scale industrial systems. Since 2019, SIM has been deployed in the display advertising system in Alibaba, bringing 7.1\% CTR and 4.4\% RPM lift, which is significant to the business. Serving the main traffic in our real system now, SIM models user behavior data with maximum length reaching up to 54000, pushing SOTA to 54x.

When discussing the threats of artificial intelligence, the first thing that comes to mind are images of Skynet, The Matrix, and the robot apocalypse. The runner up is technological unemployment, the vision of a foreseeable future in which AI algorithms take over all jobs and push humans into a struggle for meaningless survival in a world where human labor is no longer needed. Whether any or both of those threats are real is hotly debated among scientists and thought leaders. But AI algorithms also pose more imminent threats that exist today, in ways that are less conspicuous and hardly understood. In her book, Weapons of Math Destruction: How Big Data Increases Inequality and Threatens Democracy, mathematician Cathy O'Neil explores how blindly trusting algorithms to make sensitive decisions can harm many people who are on the receiving end of those decisions.

In recent years, artificial intelligence has been on the rise in the diagnosis of medical imaging. A doctor can look at an X-ray or biopsy to identify abnormalities, but this can increasingly also be done by an AI system by means of "deep learning" (see'Background: what is deep learning' below). Such a system learns to arrive at a diagnosis on its own, and in some cases it does this just as well or better than experienced doctors. The two major differences compared to a human doctor are, first, that AI is often not transparent in how it's analyzing the images, and, second, that these systems are quite "lazy." AI looks at what is needed for a particular diagnosis, and then stops.

Artificial Intelligence (AI) is impacting the world in different ways. From core industries such as manufacturing, all the way to healthcare, AI and deep learning models are changing processes and making things more efficient. And despite turbulent times, a report pinpoints that nearly half of the companies have accelerated their AI strategies, 20% doing so "significantly," betting their AI projects will have a positive impact on their organisation's resiliency, efficiency, and innovation. Yet almost half (49%) of respondents feel their company is behind in their AI journey, suggesting a critical gap exists between the strategic need and the ability to execute among business leaders and technologists. They say there is still a lack of knowledge and clarity on the applications of the technology that has been a roadblock due to which AI adoption has been slower than anticipated in the enterprise.