A Deep Dive into Adversarial Robustness in Zero-Shot Learning

2020-08-23
Yücel, Mehmet Kerim
Cinbiş, Ramazan Gökberk
Duygulu Şahin, Pınar
Machine learning (ML) systems have introduced significant advances in various fields, due to the introduction of highly complex models. Despite their success, it has been shown multiple times that machine learning models are prone to imperceptible perturbations that can severely degrade their accuracy. So far, existing studies have primarily focused on models where supervision across all classes were available. In constrast, Zero-shot Learning (ZSL) and Generalized Zero-shot Learning (GZSL) tasks inherently lack supervision across all classes. In this paper, we present a study aimed on evaluating the adversarial robustness of ZSL and GZSL models. We leverage the well-established label embedding model and subject it to a set of established adversarial attacks and defenses across multiple datasets. In addition to creating possibly the first benchmark on adversarial robustness of ZSL models, we also present analyses on important points that require attention for better interpretation of ZSL robustness results. We hope these points, along with the benchmark, will help researchers establish a better understanding what challenges lie ahead and help guide their work.

Suggestions

A mathematical contribution of statistical learning and continuous optimization using infinite and semi-infinite programming to computational statistics
Özöğür-Akyüz, Süreyya; Weber, Gerhard Wilhelm; Department of Scientific Computing (2009)
A subfield of artificial intelligence, machine learning (ML), is concerned with the development of algorithms that allow computers to “learn”. ML is the process of training a system with large number of examples, extracting rules and finding patterns in order to make predictions on new data points (examples). The most common machine learning schemes are supervised, semi-supervised, unsupervised and reinforcement learning. These schemes apply to natural language processing, search engines, medical diagnosis,...
The Effect of Loss Functions on the Deep Learning Modeling for the Flow Field Predictions Around Airfoils
Doğan, Ali; Duru, Cihat; Alemdar, Hande; Baran, Özgür Uğraş (2021-09-10)
CNNFOIL is a CNN-based machine learning tool that solves flow around the airfoil with a machine learning methodology. CNNFOIL, which is being developed by our research group, can predict flowfield around airfoils from different families at transonic regimes. We have improved the training process and accuracy of the CNNFOIL solver by implementing new loss functions. In this study, the effects of an L2 -based loss function, a physics-informed loss function based on continuity equation and a gradient differenc...
A new likelihood approach to autonomous multiple model estimation
Söken, Halil Ersin (Elsevier BV, 2020-04-01)
This paper presents an autonomous multiple model (AMM) estimation algorithm for hybrid systems with sudden changes in their parameters. Estimates of Kalman filters (KFs) that are tuned and employed for different system modes are merged based on a newly defined likelihood function without any necessity for filter interaction. The proposed likelihood function is composed of two measures, the filter agility measure and the steady-state error measure. These measures are derived based on filter adaptation rules....
Computational representation of protein sequences for homology detection and classification
Oğul, Hasan; Mumcuoğlu, Ünal Erkan; Department of Information Systems (2006)
Machine learning techniques have been widely used for classification problems in computational biology. They require that the input must be a collection of fixedlength feature vectors. Since proteins are of varying lengths, there is a need for a means of representing protein sequences by a fixed-number of features. This thesis introduces three novel methods for this purpose: n-peptide compositions with reduced alphabets, pairwise similarity scores by maximal unique matches, and pairwise similarity scores by...
A deep learning approach for the transonic flow field predictions around airfoils
Duru, Cihat; Alemdar, Hande; Baran, Özgür Uğraş (2022-01-01)
Learning from data offers new opportunities for developing computational methods in research fields, such as fluid dynamics, which constantly accumulate a large amount of data. This study presents a deep learning approach for the transonic flow field predictions around airfoils. The physics of transonic flow is integrated into the neural network model by utilizing Reynolds-averaged Navier–Stokes (RANS) simulations. A detailed investigation on the performance of the model is made both qualitatively and quant...
Citation Formats
M. K. Yücel, R. G. Cinbiş, and P. Duygulu Şahin, “A Deep Dive into Adversarial Robustness in Zero-Shot Learning,” 2020, p. 3, Accessed: 00, 2021. [Online]. Available: https://www.springer.com/gp/book/9783030664145.