Training Generative Adversarial Networks (GANs) is notoriously challenging. We propose and study an architectural modification, self-modulation, which improves GAN performance across different data sets, architectures, losses, regularizers, and hyperparameter settings. Intuitively, self-modulation allows the intermediate feature maps of a generator to change as a function of the input noise vector. While reminiscent of other conditioning techniques, it requires no labeled data. In a large-scale empirical study we observe a relative decrease of $5\%-35\%$ in FID. Furthermore, all else being equal, adding this modification to the generator leads to improved performance in $124/144$ ($86\%$) of the studied settings. Self-modulation is a simple architectural change that requires no additional parameter tuning, which suggests that it can be applied readily to any GAN.

Random Matrix Theory (RMT) is applied to analyze weight matrices of Deep Neural Networks (DNNs), including both production quality, pre-trained models such as AlexNet and Inception, and smaller models trained from scratch, such as LeNet5 and a miniature-AlexNet. Empirical and theoretical results clearly indicate that the DNN training process itself implicitly implements a form of Self-Regularization. The empirical spectral density (ESD) of DNN layer matrices displays signatures of traditionally-regularized statistical models, even in the absence of exogenously specifying traditional forms of explicit regularization. Building on relatively recent results in RMT, most notably its extension to Universality classes of Heavy-Tailed matrices, we develop a theory to identify 5+1 Phases of Training, corresponding to increasing amounts of Implicit Self-Regularization. These phases can be observed during the training process as well as in the final learned DNNs. For smaller and/or older DNNs, this Implicit Self-Regularization is like traditional Tikhonov regularization, in that there is a "size scale" separating signal from noise. For state-of-the-art DNNs, however, we identify a novel form of Heavy-Tailed Self-Regularization, similar to the self-organization seen in the statistical physics of disordered systems. This results from correlations arising at all size scales, which arises implicitly due to the training process itself. This implicit Self-Regularization can depend strongly on the many knobs of the training process. By exploiting the generalization gap phenomena, we demonstrate that we can cause a small model to exhibit all 5+1 phases of training simply by changing the batch size. This demonstrates that---all else being equal---DNN optimization with larger batch sizes leads to less-well implicitly-regularized models, and it provides an explanation for the generalization gap phenomena.

Enormous online textual information provides intriguing opportunities for understandings of social and economic semantics. In this paper, we propose a novel text regression model based on a conditional generative adversarial network (GAN), with an attempt to associate textual data and social outcomes in a semi-supervised manner. Besides promising potential of predicting capabilities, our superiorities are twofold: (i) the model works with unbalanced datasets of limited labelled data, which align with real-world scenarios; and (ii) predictions are obtained by an end-to-end framework, without explicitly selecting high-level representations. Finally we point out related datasets for experiments and future research directions.

Over the last years, the computational power of mobile devices such as smartphones and tablets has grown dramatically, reaching the level of desktop computers available not long ago. While standard smartphone apps are no longer a problem for them, there is still a group of tasks that can easily challenge even high-end devices, namely running artificial intelligence algorithms. In this paper, we present a study of the current state of deep learning in the Android ecosystem and describe available frameworks, programming models and the limitations of running AI on smartphones. We give an overview of the hardware acceleration resources available on four main mobile chipset platforms: Qualcomm, HiSilicon, MediaTek and Samsung. Additionally, we present the real-world performance results of different mobile SoCs collected with AI Benchmark that are covering all main existing hardware configurations.

For problem solving, making reactive decisions based on problem description is fast but inaccurate, while search-based planning using heuristics gives better solutions but could be exponentially slow. In this paper, we propose a new approach that improves an existing solution by iteratively picking and rewriting its local components until convergence. The rewriting policy employs a neural network trained with reinforcement learning. We evaluate our approach in two domains: job scheduling and expression simplification. Compared to common effective heuristics, baseline deep models and search algorithms, our approach efficiently gives solutions with higher quality.

For the majority of people, artificial intelligence won't be here for at least the next decade, despite all the hype. But the truth is it has already become a necessity for many businesses who work with data and it's been widely in use today. While some of us are still trying to figure out the difference between artificial intelligence and machine learning, AI is fast progressing. This breakthrough technology has already become accessible for any software developer; tech giants are currently competing to dominate the field of artificial intelligence; China has taken serious steps to become the leader in AI; some jobs might soon be automated, and we've seen some unprecedented advances in deep neural networks. AI is taking the world by storm.

With the exponential growth of the number of images (and radar data, and point clouds, and…) that are being collected, we must answer this question: how are we going to make sense of all of this data? And even before making sense of it, how are we going to sift through the amount of data to a manageable heap? How do we tell what needs further attention and what can be archived for later? The answer seems to be "Give it to the machines and let them sort it out." Now that might seem a little harsh, but it really makes sense.

A few weeks back we wrote a post on Object detection using YOLOv3. The output of an object detector is an array of bounding boxes around objects detected in the image or video frame, but we do not get any clue about the shape of the object inside the bounding box. Wouldn't it be cool if we could find a binary mask containing the object instead of just the bounding box? In this post, we will learn how to do just that. We will show how to use a Convolutional Neural Network (CNN) model called Mask-RCNN (Region based Convolutional Neural Network) for object detection and segmentation.

No Metrics Are Perfect: Adversarial Reward Learning for Visual Storytelling.

The trouble with the word artificial intelligence is the word intelligence. It misleads -- people conjure up images of thinking machines, Stephen Spielberg; Arnold Schwarzenegger coming back from the future and saying: "I'll be back." Who knows what the future may bring, but for now, and for all intents and purposes, people may be confusing intelligence and sentience. Neither are well defined -- we are supposedly sentient, machines are not, and for all we know, may never be. Intelligence means "the ability to acquire and apply knowledge and skills". These days machines can do that, machines learn, they deep learn, they can even learn by applying neural networks -- that does not make them like people, maybe they possess one subset, of a myriad of sets, that make us who we are.