As we seek to deploy machine learning models beyond virtual and controlled domains, it is critical to analyze not only the accuracy or the fact that it works most of the time, but if such a model is truly robust and reliable. This paper studies strategies to implement adversary robustly trained algorithms towards guaranteeing safety in machine learning algorithms. We provide a taxonomy to classify adversarial attacks and defenses, formulate the Robust Optimization problem in a min-max setting, and divide it into 3 subcategories, namely: Adversarial (re)Training, Regularization Approach, and Certified Defenses. We survey the most recent and important results in adversarial example generation, defense mechanisms with adversarial (re)Training as their main defense against perturbations. We also survey mothods that add regularization terms which change the behavior of the gradient, making it harder for attackers to achieve their objective. Alternatively, we've surveyed methods which formally derive certificates of robustness by exactly solving the optimization problem or by approximations using upper or lower bounds. In addition we discuss the challenges faced by most of the recent algorithms presenting future research perspectives.
Public health surveillance and tracking virus via social media can be a useful digital tool for contact tracing and preventing the spread of the virus. Nowadays, large volumes of COVID-19 tweets can quickly be processed in real-time to offer information to researchers. Nonetheless, due to the absence of labeled data for COVID-19, the preliminary supervised classifier or semi-supervised self-labeled methods will not handle non-spherical data with adequate accuracy. With the seasonal influenza and novel Coronavirus having many similar symptoms, we propose using few shot learning to fine-tune a semi-supervised model built on unlabeled COVID-19 and previously labeled influenza dataset that can provide insights into COVID-19 that have not been investigated. The experimental results show the efficacy of the proposed model with an accuracy of 86%, identification of Covid-19 related discussion using recently collected tweets.
With the ever-increasing amount of data, the central challenge in multimodal learning involves limitations of labelled samples For the task of classification, techniques such as meta-learning, zero-shot learning, and few-shot learning showcase the ability to learn information about novel classes based on prior knowledge . Recent techniques try to learn a cross-modal mapping between the semantic space and the image space. However, they tend to ignore the local and global semantic knowledge. To overcome this problem, we propose a Multimodal Variational Auto-Encoder (M-VAE) which can learn the shared latent space of image features and the semantic space. In our approach we concatenate multimodal data to a single embedding before passing it to the VAE for learning the latent space. We propose the use of a multi-modal loss during the reconstruction of the feature embedding through the decoder. Our approach is capable to correlating modalities and exploit the local and global semantic knowledge for novel sample predictions. Our experimental results using a MLP classifier on four benchmark datasets show that our proposed model outperforms the current state-of-the-art approaches for generalized zero-shot learning.
As social distancing, self-quarantines, and travel restrictions have shifted a lot of pandemic conversations to social media so does the spread of hate speech. While recent machine learning solutions for automated hate and offensive speech identification are available on Twitter, there are issues with their interpretability. We propose a novel use of learned feature importance which improves upon the performance of prior state-of-the-art text classification techniques, while producing more easily interpretable decisions. We also discuss both technical and practical challenges that remain for this task.
A key goal of cognitive neuroscience is to better understand how dynamic brain activity relates to behavior. Such dynamics, in terms of spatial and temporal patterns of brain activity, are directly measured with neurophysiological methods such as EEG, but can also be indirectly expressed by the body. Autonomic nervous system activity is the best-known example, but, muscles in the eyes and face can also index brain activity. Mostly parallel lines of artificial intelligence research show that EEG and facial muscles both encode information about emotion, pain, attention, and social interactions, among other topics. In this study, we examined adults who stutter (AWS) to understand the relations between dynamic brain and facial muscle activity and predictions about future behavior (fluent or stuttered speech). AWS can provide insight into brain-behavior dynamics because they naturally fluctuate between episodes of fluent and stuttered speech behavior. We focused on the period when speech preparation occurs, and used EEG and facial muscle activity measured from video to predict whether the upcoming speech would be fluent or stuttered. An explainable self-supervised multimodal architecture learned the temporal dynamics of both EEG and facial muscle movements during speech preparation in AWS, and predicted fluent or stuttered speech at 80.8% accuracy (chance=50%). Specific EEG and facial muscle signals distinguished fluent and stuttered trials, and systematically varied from early to late speech preparation time periods. The self-supervised architecture successfully identified multimodal activity that predicted upcoming behavior on a trial-by-trial basis. This approach could be applied to understanding the neural mechanisms driving variable behavior and symptoms in a wide range of neurological and psychiatric disorders. The combination of direct measures of neural activity and simple video data may be applied to developing technologies that estimate brain state from subtle bodily signals.
Visual Question Answering (VQA) models have achieved significant success in recent times. Despite the success of VQA models, they are mostly black-box models providing no reasoning about the predicted answer, thus raising questions for their applicability in safety-critical such as autonomous systems and cyber-security. Current state of the art fail to better complex questions and thus are unable to exploit compositionality. To minimize the black-box effect of these models and also to make them better exploit compositionality, we propose a Dynamic Neural Network (DMN), which can understand a particular question and then dynamically assemble various relatively shallow deep learning modules from a pool of modules to form a network. We incorporate compositional temporal attention to these deep learning based modules to increase compositionality exploitation. This results in achieving better understanding of complex questions and also provides reasoning as to why the module predicts a particular answer. Experimental analysis on the two benchmark datasets, VQA2.0 and CLEVR, depicts that our model outperforms the previous approaches for Visual Question Answering task as well as provides better reasoning, thus making it reliable for mission critical applications like safety and security.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.