Retrieval-augmented classification and generation models benefit from *early-stage fusion* of high-quality text-based metadata, often called memory, but face high latency and noise sensitivity. In extreme classification (XC), where low latency is crucial, existing methods use *late-stage fusion* for efficiency and robustness. To enhance accuracy while maintaining low latency, we propose MOGIC, a novel approach to metadata-infused oracle guidance for XC. We train an early-fusion oracle classifier with access to both query-side and label-side ground-truth metadata in textual form and subsequently use it to guide existing memory-based XC disciple models via regularization. The MOGIC algorithm improves precision@1 and propensity-scored precision@1 of XC disciple models by 1-2% on six standard datasets, at no additional inference-time cost. We show that MOGIC can be used in a plug-and-play manner to enhance memory-free XC models such as NGAME or DEXA. Lastly, we demonstrate the robustness of the MOGIC algorithm to missing and noisy metadata. The code is publicly available at [https://github.com/suchith720/mogic](https://github.com/suchith720/mogic).

Inverse problems form the backbone of modern signal/image processing and computational imaging, where signal reconstruction from corrupted measurements follows an optimization problem. The objective function is the sum of a data-fidelity term and a regularization functional that enforces desired properties in the reconstruction. The adversarial regularization (AR) framework is an unsupervised, data-driven approach for solving inverse problems, where the regularization function is learnt adversarially as a critique between the ground-truth distribution and the distribution of unregularized reconstructions. Thereafter, the solution to the regularized inverse problem follows an iterative technique. In this paper, we analyze the AR framework from a variational perspective, and, using Euler-Lagrange conditions, obtain the optimal regularization function in closed-form. The overall objective function is smooth and readily amenable to gradient descent minimization. We introduce momentum into the iterates as a natural extension to accelerate convergence. Since the optimal solutions are obtained in closed-form, our approach to solving inverse problems does not require prior training whilst being data-driven. We demonstrate the proposed technique on image deconvolution and show that the reconstruction performance of the proposed techniques measured in terms of peak signal-to-noise ratio (PSNR) and structural similarity index metric (SSIM) are identical to the learnt counterparts.

Score-based generative models have emerged as the state-of-the-art in generative modeling. In this paper, we introduce a novel sampling scheme that can be combined with pre-trained score-based diffusion models to speed up sampling by a factor of two to five in terms of the number of function evaluations (NFEs) with a superior Fréchet Inception distance (FID), compared to Annealed Langevin dynamics in noise-conditional score network (NCSN) and improved noise-conditional score network (NCSN++). The proposed sampling algorithm is inspired by momentum-based accelerated gradient descent used in convex optimization techniques. We validate the sampling efficiency of the proposed algorithm in terms of FID on CIFAR-10 and CelebA datasets.

We develop accurate and efficient solutions for large-scale retrieval tasks where novel (zero-shot) items can arrive continuously at a rapid pace. Conventional Siamese-style approaches embed both queries and items through a small encoder and retrieve the items lying closest to the query. While this approach allows efficient addition and retrieval of novel items, the small encoder lacks sufficient capacity for the necessary world knowledge in complex retrieval tasks. The extreme classification approaches have addressed this by learning a separate classifier for each item observed in the training set which significantly increases the representation capacity of the model. Such classifiers outperform Siamese approaches on observed items, but cannot be trained for novel items due to data and latency constraints. To bridge these gaps, this paper develops: (1) A new algorithmic framework, EMMETT, which efficiently synthesizes classifiers on-the-fly for novel items, by relying on the readily available classifiers for observed items; (2) A new algorithm, IRENE, which is a simple and effective instance of EMMETT that is specifically suited for large-scale deployments, and (3) A new theoretical framework for analyzing the generalization performance in large-scale zero-shot retrieval which guides our algorithm and training related design decisions. Comprehensive experiments are conducted on a wide range of retrieval tasks which demonstrate that IRENE improves the zero-shot retrieval accuracy by up to 15% points in Recall@10 when added on top of leading encoders. Additionally, on an online A/B test in a large-scale ad retrieval task in a major search engine, IRENE improved the ad click-through rate by 4.2%. Lastly, we validate our design choices through extensive ablative experiments. The source code for IRENE is available at https://aka.ms/irene.

Training Generative adversarial networks (GANs) stably is a challenging task. The generator in GANs transform noise vectors, typically Gaussian distributed, into realistic data such as images. In this paper, we propose a novel approach for training GANs with images as inputs, but without enforcing any pairwise constraints. The intuition is that images are more structured than noise, which the generator can leverage to learn a more robust transformation. The process can be made efficient by identifying closely related datasets, or a “friendly neighborhood” of the target distribution, inspiring the moniker, Spider GAN. To define friendly neighborhoods leveraging proximity between datasets, we propose a new measure called the signed inception distance (SID), inspired by the polyharmonic kernel. We show that the Spider GAN formulation results in faster convergence, as the generator can discover correspondence even between seemingly unrelated datasets, for instance, between TinyImageNet and CelebA faces. Further, we demonstrate cascading Spider GAN, where the output distribution from a pre-trained GAN generator is used as the input to the subsequent network. Effectively, transporting one distribution to another in a cascaded fashion until the target is learnt – a new flavor of transfer learning. We demonstrate the efficacy of the Spider approach on DCGAN, conditional GAN, PGGAN, StyleGAN2 and StyleGAN3. The proposed approach achieves state-of-the-art Fréchet inception distance (FID) values, with one-fifth of the training iterations, in comparison to their baseline counterparts on high-resolution small datasets such as MetFaces, Ukiyo-E Faces and AFHQ-Cats.

Generative adversarial networks (GANs) comprise generator and discriminator networks trained adversarially to learn the underlying distribution of a dataset. Recently, we have shown that the optimal GAN discriminator can be obtained in closed-form as the solution to the Poisson partial differential equation (PDE). While existing approaches either train a network or solve the PDE in closed-form, we propose training the generator through the gradient field of the optimal discriminator. In this paper, we establish a connection between active contour models (snakes) and GANs. We evolve a set of snake points over the gradient field of radial basis function (RBF) Coulomb GAN. The generator is then trained to follow the trajectory of the snake. The proposed approach benefits from both the sample diversity seen in flow-based approaches and the fast sampling capability of GANs. Experimental validation on 2-D synthetic data shows that the proposed approach leads to accelerated convergence, compared against the baseline approaches that either employ network or kernel-based discriminators.

We consider Generative Adversarial Networks (GANs) and address the underlying functional optimization problem ab initio within a variational setting. Strictly speaking, the optimization of the generator and discriminator functions must be carried out in accordance with the Euler-Lagrange conditions, which become particularly relevant in scenarios where the optimization cost involves regularizers comprising the derivatives of these functions. Considering Wasserstein GANs (WGANs) with a gradient-norm penalty, we show that the optimal discriminator is the solution to a Poisson differential equation. In principle, the optimal discriminator can be obtained in closed form without having to train a neural network. We illustrate this by employing a Fourier-series approximation to solve the Poisson differential equation. Experimental results based on synthesized Gaussian data demonstrate superior convergence behavior of the proposed approach in comparison with the baseline WGAN variants that employ weight-clipping, gradient or Lipschitz penalties on the discriminator on low-dimensional data. We also analyze the truncation error of the Fourier-series approximation and the estimation error of the Fourier coefficients in a high-dimensional setting. We demonstrate applications to real-world images considering latent-space prior matching in Wasserstein autoencoders and present performance comparisons on benchmark datasets such as MNIST, SVHN, CelebA, CIFAR-10, and Ukiyo-E. We demonstrate that the proposed approach achieves comparable reconstruction error and Frechet inception distance with faster convergence and up to two-fold improvement in image sharpness.

Generative adversarial networks (GANs) were originally envisioned as unsupervised generative models that learn to follow a target distribution. Variants such as conditional GANs, auxiliary-classifier GANs (ACGANs) project GANs on to supervised and semi-supervised learning frameworks by providing labelled data and using multi-class discriminators. In this paper, we approach the supervised GAN problem from a different perspective, one that is motivated by the philosophy of the famous Persian poet Rumi who said, "The art of knowing is knowing what to ignore." In the GAN framework, we not only provide the GAN positive data that it must learn to model, but also present it with so-called negative samples that it must learn to avoid — we call this "The Rumi Framework." This formulation allows the discriminator to represent the underlying target distribution better by learning to penalize generated samples that are undesirable — we show that this capability accelerates the learning process of the generator. We present a reformulation of the standard GAN (SGAN) and least-squares GAN (LSGAN) within the Rumi setting. The advantage of the reformulation is demonstrated by means of experiments conducted on MNIST, Fashion MNIST, CelebA, and CIFAR-10 datasets. Finally, we consider an application of the proposed formulation to address the important problem of learning an under-represented class in an unbalanced dataset. The Rumi approach results in substantially lower FID scores than the standard GAN frameworks while possessing better generalization capability.