Deep-Learning-as-a-Service, unveiled at IBM's annual IT industry conference in Las Vegas, seeks to lower barriers to deploying AI and deep-learning tools, a complex and painstakingly repetitive process that requires large amounts of computing power, the company said. The new service allows companies to upload data in Watson Studio, IBM's cloud-native platform for data scientists, developers and business analysts. There, they can create deep-learning algorithms for datasets – known in AI parlance as a "neural network" – using a drag-and-drop interface to select, configure, design and code the network. IBM also has automated the repetitive process of fine-tuning deep-learning algorithms, with successive training runs started, monitored and stopped automatically. For many firms, the complexity of creating smart algorithms from scratch has kept them from leveraging AI to parse massive stores of data for business value, the company said.

We present a new approach to learning for planning, where knowledge acquired while solving a given set of planning problems is used to plan faster in related, but new problem instances. We show that a deep neural network can be used to learn and represent a generalized reactive policy (GRP) that maps a problem instance and a state to an action, and that the learned GRPs efficiently solve large classes of challenging problem instances. In contrast to prior efforts in this direction, our approach significantly reduces the dependence of learning on handcrafted domain knowledge or feature selection. Instead, the GRP is trained from scratch using a set of successful execution traces. We show that our approach can also be used to automatically learn a heuristic function that can be used in directed search algorithms. We evaluate our approach using an extensive suite of experiments on two challenging planning problem domains and show that our approach facilitates learning complex decision making policies and powerful heuristic functions with minimal human input

Adversarial attack methods have demonstrated the fragility of deep neural networks. Their imperceptible perturbations are frequently able fool classifiers into potentially dangerous misclassifications. We propose a novel way to interpret adversarial perturbations in terms of the effective input signal that classifiers actually use. Based on this, we apply specially trained autoencoders, referred to as S2SNets, as defense mechanism. They follow a two-stage training scheme: first unsupervised, followed by a fine-tuning of the decoder, using gradients from an existing classifier. S2SNets induce a shift in the distribution of gradients propagated through them, stripping them from class-dependent signal. We analyze their robustness against several white-box and gray-box scenarios on the large ImageNet dataset. Our approach reaches comparable resilience in white-box attack scenarios as other state-of-the-art defenses in gray-box scenarios. We further analyze the relationships of AlexNet, VGG 16, ResNet 50 and Inception v3 in adversarial space, and found that VGG 16 is the easiest to fool, while perturbations from ResNet 50 are the most transferable.

Considered as a data-driven approach, Fingerprinting Localization Solutions (FPSs) enjoy huge popularity due to their good performance and minimal environment information requirement. This papers addresses applications of artificial intelligence to solve two problems in Received Signal Strength Indicator (RSSI) based FPS, first the cumbersome training database construction and second the extrapolation of fingerprinting algorithm for similar buildings with slight environmental changes. After a concise overview of deep learning design techniques, two main techniques widely used in deep learning are exploited for the above mentioned issues namely data augmentation and transfer learning. We train a multi-layer neural network that learns the mapping from the observations to the locations. A data augmentation method is proposed to increase the training database size based on the structure of RSSI measurements and hence reducing effectively the amount of training data. Then it is shown experimentally how a model trained for a particular building can be transferred to a similar one by fine tuning with significantly smaller training numbers. The paper implicitly discusses the new guidelines to consider about deep learning designs when they are employed in a new application context.

Information planning enables faster learning with fewer training examples. It is particularly applicable when training examples are costly to obtain. This work examines the advantages of information planning for text data by focusing on three supervised models: Naive Bayes, supervised LDA and deep neural networks. We show that planning based on entropy and mutual information outperforms random selection baseline and therefore accelerates learning.

In this work we explore a straightforward variational Bayes scheme for Recurrent Neural Networks. Firstly, we show that a simple adaptation of truncated backpropagation through time can yield good quality uncertainty estimates and superior regularisation at only a small extra computational cost during training, also reducing the amount of parameters by 80\%. Secondly, we demonstrate how a novel kind of posterior approximation yields further improvements to the performance of Bayesian RNNs. We incorporate local gradient information into the approximate posterior to sharpen it around the current batch statistics. We show how this technique is not exclusive to recurrent neural networks and can be applied more widely to train Bayesian neural networks. We also empirically demonstrate how Bayesian RNNs are superior to traditional RNNs on a language modelling benchmark and an image captioning task, as well as showing how each of these methods improve our model over a variety of other schemes for training them. We also introduce a new benchmark for studying uncertainty for language models so future methods can be easily compared.

Cryo-electron microscopy (cryoEM) is fast becoming the preferred method for protein structure determination. Particle picking is a significant bottleneck in the solving of protein structures from single particle cryoEM. Hand labeling sufficient numbers of particles can take months of effort and current computationally based approaches are often ineffective. Here, we frame particle picking as a positive-unlabeled classification problem in which we seek to learn a convolutional neural network (CNN) to classify micrograph regions as particle or background from a small number of labeled positive examples and many unlabeled examples. However, model fitting with very few labeled data points is a challenging machine learning problem. To address this, we develop a novel objective function, GE-binomial, for learning model parameters in this context. This objective uses a newly-formulated generalized expectation criteria to learn effectively from unlabeled data when using minibatched stochastic gradient descent optimizers. On a high-quality publicly available cryoEM dataset and a difficult unpublished dataset supplied by the Shapiro lab, we show that CNNs trained with this objective classify particles accurately with very few positive training examples and outperform EMAN2's byRef method by a large margin even with fewer labeled training examples. Furthermore, we show that incorporating an autoencoder improves generalization when very few labeled data points are available. We also compare our GE-binomial method with other positive-unlabeled learning methods never before applied to particle picking. We expect our particle picking tool, Topaz, based on CNNs trained with GE-binomial, to be an essential component of single particle cryoEM analysis and our GE-binomial objective function to be widely applicable to positive-unlabeled classification problems.

Generative Adversarial Networks (GANs) are becoming popular choices for unsupervised learning. At the same time there is a concerted effort in the machine learning community to expand the range of tasks in which learning can be applied as well as to utilize methods from other disciplines to accelerate learning. With this in mind, in the current work we suggest ways to enforce given constraints in the output of a GAN both for interpolation and extrapolation. The two cases need to be treated differently. For the case of interpolation, the incorporation of constraints is built into the training of the GAN. The incorporation of the constraints respects the primary game-theoretic setup of a GAN so it can be combined with existing algorithms. However, it can exacerbate the problem of instability during training that is well-known for GANs. We suggest adding small noise to the constraints as a simple remedy that has performed well in our numerical experiments. The case of extrapolation (prediction) is more involved. First, we employ a modified interpolation training process that uses noisy data but does not necessarily enforce the constraints during training. Second, the resulting modified interpolator is used for extrapolation where the constraints are enforced after each step through projection on the space of constraints.

We revisit the Blind Deconvolution problem with a focus on understanding its robustness and convergence properties. Provable robustness to noise and other perturbations is receiving recent interest in vision, from obtaining immunity to adversarial attacks to assessing and describing failure modes of algorithms in mission critical applications. Further, many blind deconvolution methods based on deep architectures internally make use of or optimize the basic formulation, so a clearer understanding of how this sub-module behaves, when it can be solved, and what noise injection it can tolerate is a first order requirement. We derive new insights into the theoretical underpinnings of blind deconvolution. The algorithm that emerges has nice convergence guarantees and is provably robust in a sense we formalize in the paper. Interestingly, these technical results play out very well in practice, where on standard datasets our algorithm yields results competitive with or superior to the state of the art.

Modern machine learning techniques, such as convolutional, recurrent and recursive neural networks, have shown promise for jet substructure at the Large Hadron Collider. For example, they have demonstrated effectiveness at boosted top or W boson identification or for quark/gluon discrimination. We explore these methods for the purpose of classifying jets according to their electric charge. We find that neural networks that incorporate distance within the jet as an input can provide significant improvement in jet charge extraction over traditional methods. We find that both convolutional and recurrent networks are effective and both train faster than recursive networks. The advantages of using a fixed-size input representation (as with the CNN) or a smaller input representation (as with the RNN) suggest that both convolutional and recurrent networks will be essential to the future of modern machine learning at colliders.