It is well established that increasing scale in deep transformer networks leads to improved quality and performance. This increase in scale often comes with an increase in compute cost and inference latency. Consequently, research into methods which help realize the benefits of increased scale without leading to an increase in the compute cost becomes important. We introduce Alternating Updates (AltUp), a simple-to-implement method to increase a model's capacity without the computational burden. AltUp enables the widening of the learned representation without increasing the computation time by working on a subblock of the representation at each layer. Our experiments on various transformer models and language tasks demonstrate the consistent effectiveness of alternating updates on a diverse set of benchmarks. Finally, we present extensions of AltUp to the sequence dimension, and demonstrate how AltUp can be synergistically combined with existing approaches, such as Sparse Mixture-of-Experts models, to obtain efficient models with even higher capacity.
Deep neural networks have been demonstrated to be vulnerable to adversarial attacks: subtle perturbation can completely change the prediction result. Existing adversarial attacks on object detection focus on attacking anchor-based detectors, which may not work well for anchor-free detectors. In this paper, we propose the first adversarial attack dedicated to anchor-free detectors. It is a category-wise attack that attacks important pixels of all instances of a category simultaneously. Our attack manifests in two forms, sparse category-wise attack (SCA) and dense category-wise attack (DCA), that minimize the $L_0$ and $L_\infty$ norm-based perturbations, respectively. For DCA, we present three variants, DCA-G, DCA-L, and DCA-S, that select a global region, a local region, and a semantic region, respectively, to attack. Our experiments on large-scale benchmark datasets including PascalVOC, MS-COCO, and MS-COCO Keypoints indicate that our proposed methods achieve state-of-the-art attack performance and transferability on both object detection and human pose estimation tasks.
Reconfigurable intelligent surfaces (RISs) can potentially combat jamming attacks by diffusing jamming signals. This paper jointly optimizes user selection, channel allocation, modulation-coding, and RIS configuration in a multiuser OFDMA system under a jamming attack. This problem is non-trivial and has never been addressed, because of its mixed-integer programming nature and difficulties in acquiring channel state information (CSI) involving the RIS and jammer. We propose a new deep reinforcement learning (DRL)-based approach, which learns only through changes in the received data rates of the users to reject the jamming signals and maximize the sum rate of the system. The key idea is that we decouple the discrete selection of users, channels, and modulation-coding from the continuous RIS configuration, hence facilitating the RIS configuration with the latest twin delayed deep deterministic policy gradient (TD3) model. Another important aspect is that we show a winner-takes-all strategy is almost surely optimal for selecting the users, channels, and modulation-coding, given a learned RIS configuration. Simulations show that the new approach converges fast to fulfill the benefit of the RIS, due to its substantially small state and action spaces. Without the need of the CSI, the approach is promising and offers practical value.
Pretrained large-scale vision-language models like CLIP have exhibited strong generalization over unseen tasks. Yet imperceptible adversarial perturbations can significantly reduce CLIP's performance on new tasks. In this work, we identify and explore the problem of \emph{adapting large-scale models for zero-shot adversarial robustness}. We first identify two key factors during model adaption -- training losses and adaptation methods -- that affect the model's zero-shot adversarial robustness. We then propose a text-guided contrastive adversarial training loss, which aligns the text embeddings and the adversarial visual features with contrastive learning on a small set of training data. We apply this training loss to two adaption methods, model finetuning and visual prompt tuning. We find that visual prompt tuning is more effective in the absence of texts, while finetuning wins in the existence of text guidance. Overall, our approach significantly improves the zero-shot adversarial robustness over CLIP, seeing an average improvement of over 31 points over ImageNet and 15 zero-shot datasets. We hope this work can shed light on understanding the zero-shot adversarial robustness of large-scale models.
Many visual recognition models are evaluated only on their classification accuracy, a metric for which they obtain strong performance. In this paper, we investigate whether computer vision models can also provide correct rationales for their predictions. We propose a ``doubly right'' object recognition benchmark, where the metric requires the model to simultaneously produce both the right labels as well as the right rationales. We find that state-of-the-art visual models, such as CLIP, often provide incorrect rationales for their categorical predictions. However, by transferring the rationales from language models into visual representations through a tailored dataset, we show that we can learn a ``why prompt,'' which adapts large visual representations to produce correct rationales. Visualizations and empirical experiments show that our prompts significantly improve performance on doubly right object recognition, in addition to zero-shot transfer to unseen tasks and datasets.
The use of modern vocoders in an analysis/synthesis pipeline allows us to investigate high-quality voice conversion that can be used for privacy purposes. Here, we propose to transform the speaker embedding and the pitch in order to hide the sex of the speaker. ECAPA-TDNN-based speaker representation fed into a HiFiGAN vocoder is protected using a neural-discriminant analysis approach, which is consistent with the zero-evidence concept of privacy. This approach significantly reduces the information in speech related to the speaker's sex while preserving speech content and some consistency in the resulting protected voices.
In recent years, Deep Learning has shown good results in the Single Image Superresolution Reconstruction (SISR) task, thus becoming the most widely used methods in this field. The SISR task is a typical task to solve an uncertainty problem. Therefore, it is often challenging to meet the requirements of High-quality sampling, fast Sampling, and diversity of details and texture after Sampling simultaneously in a SISR task.It leads to model collapse, lack of details and texture features after Sampling, and too long Sampling time in High Resolution (HR) image reconstruction methods. This paper proposes a Diffusion Probability model for Latent features (LDDPM) to solve these problems. Firstly, a Conditional Encoder is designed to effectively encode Low-Resolution (LR) images, thereby reducing the solution space of reconstructed images to improve the performance of reconstructed images. Then, the Normalized Flow and Multi-modal adversarial training are used to model the denoising distribution with complex Multi-modal distribution so that the Generative Modeling ability of the model can be improved with a small number of Sampling steps. Experimental results on mainstream datasets demonstrate that our proposed model reconstructs more realistic HR images and obtains better PSNR and SSIM performance compared to existing SISR tasks, thus providing a new idea for SISR tasks.
With the similarity between music and speech synthesis from symbolic input and the rapid development of text-to-speech (TTS) techniques, it is worthwhile to explore ways to improve the MIDI-to-audio performance by borrowing from TTS techniques. In this study, we analyze the shortcomings of a TTS-based MIDI-to-audio system and improve it in terms of feature computation, model selection, and training strategy, aiming to synthesize highly natural-sounding audio. Moreover, we conducted an extensive model evaluation through listening tests, pitch measurement, and spectrogram analysis. This work demonstrates not only synthesis of highly natural music but offers a thorough analytical approach and useful outcomes for the community. Our code and pre-trained models are open sourced at https://github.com/nii-yamagishilab/midi-to-audio.
Most existing domain adaptation (DA) methods align the features based on the domain feature distributions and ignore aspects related to fog, background and target objects, rendering suboptimal performance. In our DA framework, we retain the depth and background information during the domain feature alignment. A consistency loss between the generated depth and fog transmission map is introduced to strengthen the retention of the depth information in the aligned features. To address false object features potentially generated during the DA process, we propose an encoder-decoder framework to reconstruct the fog-free background image. This reconstruction loss also reinforces the encoder, i.e., our DA backbone, to minimize false object features.Moreover, we involve our target data in training both our DA module and our detection module in a semi-supervised manner, so that our detection module is also exposed to the unlabeled target data, the type of data used in the testing stage. Using these ideas, our method significantly outperforms the state-of-the-art method (47.6 mAP against the 44.3 mAP on the Foggy Cityscapes dataset), and obtains the best performance on multiple real-image public datasets. Code is available at: https://github.com/VIML-CVDL/Object-Detection-in-Foggy-Scenes
Disentangled Representation Learning (DRL) aims to learn a model capable of identifying and disentangling the underlying factors hidden in the observable data in representation form. The process of separating underlying factors of variation into variables with semantic meaning benefits in learning explainable representations of data, which imitates the meaningful understanding process of humans when observing an object or relation. As a general learning strategy, DRL has demonstrated its power in improving the model explainability, controlability, robustness, as well as generalization capacity in a wide range of scenarios such as computer vision, natural language processing, data mining etc. In this article, we comprehensively review DRL from various aspects including motivations, definitions, methodologies, evaluations, applications and model designs. We discuss works on DRL based on two well-recognized definitions, i.e., Intuitive Definition and Group Theory Definition. We further categorize the methodologies for DRL into four groups, i.e., Traditional Statistical Approaches, Variational Auto-encoder Based Approaches, Generative Adversarial Networks Based Approaches, Hierarchical Approaches and Other Approaches. We also analyze principles to design different DRL models that may benefit different tasks in practical applications. Finally, we point out challenges in DRL as well as potential research directions deserving future investigations. We believe this work may provide insights for promoting the DRL research in the community.