It can be difficult to assess the quality of a fitted model when facing unsupervised learning problems. Latent variable models, such as variation autoencoders and Gaussian mixture models, are often trained with likelihood-based approaches. In scope of Goodhart's law, when a metric becomes a target it ceases to be a good metric and therefore we should not use likelihood to assess the quality of the fit of these models. The solution we propose is a new metric for model comparison or regularization that relies on moments. The concept is to study the difference between the data moments and the model moments using a matrix norm, such as the Frobenius norm. We show how to use this new metric for model comparison and then for regularization. It is common to draw samples from the fitted distribution when evaluating latent variable models and we show that our proposed metric is faster to compute and has a smaller variance that this alternative. We conclude this article with a proof of concept of both applications and we discuss future work.
相關內容
Deep generative models are a prominent approach for data generation, and have been used to produce high quality samples in various domains. Diffusion models, an emerging class of deep generative models, have attracted considerable attention owing to their exceptional generative quality. Despite this, they have certain limitations, including a time-consuming iterative generation process and confinement to high-dimensional Euclidean space. This survey presents a plethora of advanced techniques aimed at enhancing diffusion models, including sampling acceleration and the design of new diffusion processes. In addition, we delve into strategies for implementing diffusion models in manifold and discrete spaces, maximum likelihood training for diffusion models, and methods for creating bridges between two arbitrary distributions. The innovations we discuss represent the efforts for improving the functionality and efficiency of diffusion models in recent years. To examine the efficacy of existing models, a benchmark of FID score, IS, and NLL is presented in a specific NFE. Furthermore, diffusion models are found to be useful in various domains such as computer vision, audio, sequence modeling, and AI for science. The paper concludes with a summary of this field, along with existing limitations and future directions. Summation of existing well-classified methods is in our Github: //github.com/chq1155/A-Survey-on-Generative-Diffusion-Model
Synthetic data generation has been a growing area of research in recent years. However, its potential applications in serious games have not been thoroughly explored. Advances in this field could anticipate data modelling and analysis, as well as speed up the development process. To try to fill this gap in the literature, we propose a simulator architecture for generating probabilistic synthetic data for serious games based on interactive narratives. This architecture is designed to be generic and modular so that it can be used by other researchers on similar problems. To simulate the interaction of synthetic players with questions, we use a cognitive testing model based on the Item Response Theory framework. We also show how probabilistic graphical models (in particular Bayesian networks) can be used to introduce expert knowledge and external data into the simulation. Finally, we apply the proposed architecture and methods in a use case of a serious game focused on cyberbullying. We perform Bayesian inference experiments using a hierarchical model to demonstrate the identifiability and robustness of the generated data.
Despite the significant recent progress in deep generative models, the underlying structure of their latent spaces is still poorly understood, thereby making the task of performing semantically meaningful latent traversals an open research challenge. Most prior work has aimed to solve this challenge by modeling latent structures linearly, and finding corresponding linear directions which result in `disentangled' generations. In this work, we instead propose to model latent structures with a learned dynamic potential landscape, thereby performing latent traversals as the flow of samples down the landscape's gradient. Inspired by physics, optimal transport, and neuroscience, these potential landscapes are learned as physically realistic partial differential equations, thereby allowing them to flexibly vary over both space and time. To achieve disentanglement, multiple potentials are learned simultaneously, and are constrained by a classifier to be distinct and semantically self-consistent. Experimentally, we demonstrate that our method achieves both more qualitatively and quantitatively disentangled trajectories than state-of-the-art baselines. Further, we demonstrate that our method can be integrated as a regularization term during training, thereby acting as an inductive bias towards the learning of structured representations, ultimately improving model likelihood on similarly structured data.
We investigate a general matrix factorization for deviance-based data losses, extending the ubiquitous singular value decomposition beyond squared error loss. While similar approaches have been explored before, our method leverages classical statistical methodology from generalized linear models (GLMs) and provides an efficient algorithm that is flexible enough to allow for structural zeros via entry weights. Moreover, by adapting results from GLM theory, we provide support for these decompositions by (i) showing strong consistency under the GLM setup, (ii) checking the adequacy of a chosen exponential family via a generalized Hosmer-Lemeshow test, and (iii) determining the rank of the decomposition via a maximum eigenvalue gap method. To further support our findings, we conduct simulation studies to assess robustness to decomposition assumptions and extensive case studies using benchmark datasets from image face recognition, natural language processing, network analysis, and biomedical studies. Our theoretical and empirical results indicate that the proposed decomposition is more flexible, general, and robust, and can thus provide improved performance when compared to similar methods. To facilitate applications, an R package with efficient model fitting and family and rank determination is also provided.
Detecting fake images is becoming a major goal of computer vision. This need is becoming more and more pressing with the continuous improvement of synthesis methods based on Generative Adversarial Networks (GAN), and even more with the appearance of powerful methods based on Diffusion Models (DM). Towards this end, it is important to gain insight into which image features better discriminate fake images from real ones. In this paper we report on our systematic study of a large number of image generators of different families, aimed at discovering the most forensically relevant characteristics of real and generated images. Our experiments provide a number of interesting observations and shed light on some intriguing properties of synthetic images: (1) not only the GAN models but also the DM and VQ-GAN (Vector Quantized Generative Adversarial Networks) models give rise to visible artifacts in the Fourier domain and exhibit anomalous regular patterns in the autocorrelation; (2) when the dataset used to train the model lacks sufficient variety, its biases can be transferred to the generated images; (3) synthetic and real images exhibit significant differences in the mid-high frequency signal content, observable in their radial and angular spectral power distributions.
Classifiers based on deep neural networks have been recently challenged by Adversarial Attack, where the widely existing vulnerability has invoked the research in defending them from potential threats. Given a vulnerable classifier, existing defense methods are mostly white-box and often require re-training the victim under modified loss functions/training regimes. While the model/data/training specifics of the victim are usually unavailable to the user, re-training is unappealing, if not impossible for reasons such as limited computational resources. To this end, we propose a new black-box defense framework. It can turn any pre-trained classifier into a resilient one with little knowledge of the model specifics. This is achieved by new joint Bayesian treatments on the clean data, the adversarial examples and the classifier, for maximizing their joint probability. It is further equipped with a new post-train strategy which keeps the victim intact. We name our framework Bayesian Boundary Correction (BBC). BBC is a general and flexible framework that can easily adapt to different data types. We instantiate BBC for image classification and skeleton-based human activity recognition, for both static and dynamic data. Exhaustive evaluation shows that BBC has superior robustness and can enhance robustness without severely hurting the clean accuracy, compared with existing defense methods.
Our goal is to produce methods for observational causal inference that are auditable, easy to troubleshoot, accurate for treatment effect estimation, and scalable to high-dimensional data. We describe a general framework called Model-to-Match that achieves these goals by (i) learning a distance metric via outcome modeling, (ii) creating matched groups using the distance metric, and (iii) using the matched groups to estimate treatment effects. Model-to-Match uses variable importance measurements to construct a distance metric, making it a flexible framework that can be adapted to various applications. Concentrating on the scalability of the problem in the number of potential confounders, we operationalize the Model-to-Match framework with LASSO. We derive performance guarantees for settings where LASSO outcome modeling consistently identifies all confounders (importantly without requiring the linear model to be correctly specified). We also provide experimental results demonstrating the method's auditability, accuracy, and scalability as well as extensions to more general nonparametric outcome modeling.
This paper outlines an end-to-end optimized lossy image compression framework using diffusion generative models. The approach relies on the transform coding paradigm, where an image is mapped into a latent space for entropy coding and, from there, mapped back to the data space for reconstruction. In contrast to VAE-based neural compression, where the (mean) decoder is a deterministic neural network, our decoder is a conditional diffusion model. Our approach thus introduces an additional "content" latent variable on which the reverse diffusion process is conditioned and uses this variable to store information about the image. The remaining "texture" variables characterizing the diffusion process are synthesized at decoding time. We show that the model's performance can be tuned toward perceptual metrics of interest. Our extensive experiments involving multiple datasets and image quality assessment metrics show that our approach yields stronger reported FID scores than the GAN-based model, while also yielding competitive performance with VAE-based models in several distortion metrics. Furthermore, training the diffusion with X-parameterization enables high-quality reconstructions in only a handful of decoding steps, greatly affecting the model's practicality.
Since hardware resources are limited, the objective of training deep learning models is typically to maximize accuracy subject to the time and memory constraints of training and inference. We study the impact of model size in this setting, focusing on Transformer models for NLP tasks that are limited by compute: self-supervised pretraining and high-resource machine translation. We first show that even though smaller Transformer models execute faster per iteration, wider and deeper models converge in significantly fewer steps. Moreover, this acceleration in convergence typically outpaces the additional computational overhead of using larger models. Therefore, the most compute-efficient training strategy is to counterintuitively train extremely large models but stop after a small number of iterations. This leads to an apparent trade-off between the training efficiency of large Transformer models and the inference efficiency of small Transformer models. However, we show that large models are more robust to compression techniques such as quantization and pruning than small models. Consequently, one can get the best of both worlds: heavily compressed, large models achieve higher accuracy than lightly compressed, small models.
With the rapid increase of large-scale, real-world datasets, it becomes critical to address the problem of long-tailed data distribution (i.e., a few classes account for most of the data, while most classes are under-represented). Existing solutions typically adopt class re-balancing strategies such as re-sampling and re-weighting based on the number of observations for each class. In this work, we argue that as the number of samples increases, the additional benefit of a newly added data point will diminish. We introduce a novel theoretical framework to measure data overlap by associating with each sample a small neighboring region rather than a single point. The effective number of samples is defined as the volume of samples and can be calculated by a simple formula $(1-\beta^{n})/(1-\beta)$, where $n$ is the number of samples and $\beta \in [0,1)$ is a hyperparameter. We design a re-weighting scheme that uses the effective number of samples for each class to re-balance the loss, thereby yielding a class-balanced loss. Comprehensive experiments are conducted on artificially induced long-tailed CIFAR datasets and large-scale datasets including ImageNet and iNaturalist. Our results show that when trained with the proposed class-balanced loss, the network is able to achieve significant performance gains on long-tailed datasets.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191