Symmetric orthogonalization via SVD, and closely related procedures, are well-known techniques for projecting matrices onto O(n) or SO(n). These tools have long been used for applications in computer vision, for example optimal 3D alignment problems solved by orthogonal Procrustes, rotation averaging, or Essential matrix decomposition. Despite its utility in different settings, SVD orthogonalization as a procedure for producing rotation matrices is typically overlooked in deep learning models, where the preferences tend toward classic representations like unit quaternions, Euler angles, and axis-angle, or more recently-introduced methods. Despite the importance of 3D rotations in computer vision and robotics, a single universally effective representation is still missing. Here, we explore the viability of SVD orthogonalization for 3D rotations in neural networks.

Generative Adversarial Networks (GANs) produce high-quality images but are challenging to train. They need careful regularization, vast amounts of compute, and expensive hyper-parameter sweeps. We make significant headway on these issues by projecting generated and real samples into a fixed, pretrained feature space. Motivated by the finding that the discriminator cannot fully exploit features from deeper layers of the pretrained model, we propose a more effective strategy that mixes features across channels and resolutions. Our Projected GAN improves image quality, sample efficiency, and convergence speed.

In modern multilabel classification problems, each data instance belongs to a small number of classes among a large set of classes. In other words, these problems involve learning very sparse binary label vectors. Moreover, in the large-scale problems, the labels typically have certain (unknown) hierarchy. In this paper we exploit the sparsity of label vectors and the hierarchical structure to embed them in low-dimensional space using label groupings. Consequently, we solve the classification problem in a much lower dimensional space and then obtain labels in the original space using an appropriately defined lifting.

Does it matter whether one randomly initializes a neural network (NN) from Gaussian, uniform, or other distributions? We show the answer is "yes" in some parameter tensors (the so-called matrix-like parameters) but "no" in others when the NN is wide. This is a specific instance of a more general universality principle for Tensor Programs (TP) that informs precisely when the limit of a program depends on the distribution of its initial matrices and vectors. To obtain this principle, we develop the theory of non-Gaussian Tensor Programs. As corollaries, we obtain all previous consequences of the TP framework (such as NNGP/NTK correspondence, Free Independence Principle, Dynamical Dichotomy Theorem, and μ-parametrization) for NNs with non-Gaussian weights.

We develop a new exponential family probabilistic model for permutations that can capture hierarchical structure, and that has the well known Mallows and generalized Mallows models as subclasses. We describe how one can do parameter estimation and propose an approach to structure search for this class of models. We provide experimental evidence that this added flexibility both improves predictive performance and enables a deeper understanding of collections of permutations.

Squid Game is not a subtle show. It is impossible to misinterpret its very obvious message that the games are bad, and people should NOT be driven to such desperation by a merciless capitalist system that they will murder each other for rich people's entertainment. I would not be the first to point out that there is some conflict in the fact that we, the viewers, are watching all these competitors get killed for our entertainment, but still: despite the violence, despite the shock value, there is no ambiguity around the narrative intention. In this spin-off video game from Netflix, by contrast, the games are not bad. They are supposed to be fun.

TL;DR: With VideoProc, AI is your video editor, and it's only 29.97 for a lifetime license when you use promo code PROCLIFETIME through February 2. It's so annoying when you see something amazing and actually manage to catch it on video, only to find out your video is blurry, shaky, and totally unwatchable. Good thing there's an AI tool that can fix that. VideoProc Converter AI is a tool that lets you enhance your videos in a matter of moments, and you don't need to be an AI expert to use it. The video of baby's first steps turned out blurry, but don't worry. VideoProc lets you upscale videos to 4K and preserve every precious detail.

Jon Voight spoke to Fox News Digital while promoting his film "Reagan" and weighed in on the family drama between his daughter, Angelina Jolie, and her ex, Brad Pitt. A happily-ever-after with whom a woman assumed to be Hollywood hunk Brad Pitt quickly turned into a living nightmare. On Jan. 12, the French television channel TF1 aired an episode of its show "Sept à Huit," which told the story of a 53-year-old interior designer named Anne who revealed that she had lost 830,000 euros (approximately 850,000) in personal funds because she thought she was sending money to a cancer-ridden Pitt. Through falsified documents and images as well as artificial intelligence, Anne believed she was speaking to, and eventually in a relationship with, the 61-year-old actor. WHAT IS ARTIFICIAL INTELLIGENCE (AI)? A woman was conned into believing she was in a relationship with Brad Pitt after being contacted by someone claiming to be the actor on Instagram.

Keir Starmer this week announced a 50-point plan that aims to give the UK world leader status in artificial intelligence and grow the economy by as much as 47bn a year over a decade. The multibillion-pound investment, which seeks to create a 20-fold increase in the amount of AI computing power under public control by 2030, has been framed as a gamechanger for businesses and public organisations. The reaction to the announcement has been mixed, given it is far from clear that the much-hyped potential of AI will result in the level of economic benefit forecast. Many are concerned that the technology could lead to widespread job cuts, while others fear a destruction in the value and growth of the creative industries after learning of proposals to make it easier for AI companies to mine artistic works for data, for no cost. Despite such concerns, for many in the world of business the AI revolution is already here and transforming their industries.

Deploying deep convolutional neural networks on Internet-of-Things (IoT) devices is challenging due to the limited computational resources, such as limited SRAM memory and Flash storage. Previous works re-design a small network for IoT devices, and then compress the network size by mixed-precision quantization. In this work, we propose a one-stage solution that optimizes both jointly and automatically. The key idea of our approach is to cast the joint architecture design and quantization as an Entropy Maximization process. Particularly, our algorithm automatically designs a tiny deep model such that: 1) Its representation capacity measured by entropy is maximized under the given computational budget; 2) Each layer is assigned with a proper quantization precision; 3) The overall design loop can be done on CPU, and no GPU is required.