Convex Learning with Invariances

By Choon Hui Teo, Amir Globerson, Sam T. Roweis, Alex Smola. 2007

114 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Statistical Learning Theory

Citations in the corpus

Poisoning Attacks against Support Vector Machines (2012, methodology): “In response to the threat of adversarial data manipulation, several proposed learning methods explicitly account for certain types of corrupted data (Globerson & Roweis, 2006; Teo et al., 2008; Br¨uckner & Schef-fer, 2009; Dekel et al., 2010).”
Is Feature Selection Secure against Training Data Poisoning? (2015, background): “In several cases, these constraints can be encoded in terms of distance measures between the original, non-manipulated attack samples and the manipulated ones (Dalvi et al., 2004; Lowd & Meek, 2005; Globerson & Roweis, 2006; Teo et al., 2008; Br ¨ uckner et al., 2012; Biggio et al., 2013a).”
Adversarial Detection of Flash Malware: Limitations and Open Issues (2017, background): “In many cases, these constraints can be directly encoded in terms of distances in feature space, computed between the source malware data and its manipulated versions [8], [13], [14], [16], [21], [29].”
Yes, Machine Learning Can Be More Secure! A Case Study on Android Malware Detection (2017, background): “puted between the source malware data and its manipulated versions [7], [15], [19], [21], [29], [40].”
Adversarial Feature Selection Against Evasion Attacks (2016, background): “theoretical [8], [12], [20], [30] or probabilistic models of the hypothesized attack strategy [2], [9].”
Support vector machines under adversarial label contamination (2015, methodology): “, using game-theoretical models for classification [15, 20–22], probabilistic models of the data distribution drift under attack [23, 24], and even multiple classifier systems [25–27].”
Robustness and Regularization of Support Vector Machines (2008, background): “our reasons that motivate this “robust to perturbation” setup and in particular the min-max form of (1) and (2). First, it can explicitly incorporate prior problem knowledge of local invariance (e.g, Teo et al., 2008). For example, in vision tasks, a desirable classiﬁer should provide a consistent a”
Learning to Compose Domain-Specific Transformations for Data Augmentation (2017, methodology): “In one set of approaches, transformation operations are selected adaptively from a given set in order to maximize the loss of the end classification model being trained [30, 8].”
Identifying Unknown Unknowns in the Open World: Representations and Policies for Guided Exploration (2016, methodology): “Multiple approaches have been proposed to address dataset shift, including importance weighting of training instances based on similarity to test set (Shimodaira 2000), online learning of prediction models (Cesa-Bianchi and Lugosi 2006), and learning models robust to adversarial actions (Teo et al. ”
One-and-a-Half-Class Multiple Classifier Systems for Secure Learning Against Evasion Attacks at Test Time (2015, background): “Compared to other secure learning techniques [12,13,14], we do not make any specific assumption on the kind of adversarial drift that may occur at test time, but rather only agnostically assume that malicious data may appear everywhere in feature space at test time with a non-null probability.”
On Robustness and Regularization of Structural Support Vector Machines (2014, background): “Teo et al. (2008) extend this to any adversarial manipulation that can be efficiently simulated.”
Secure Kernel Machines against Evasion Attacks (2016, background): “The underling idea of the work in [14, 31] is instead to consider a worst-case loss suited to sparse attacks in which features can be set to their maximum/minimum values.”
Support Vector Machines Under Adversarial Label Noise (2011, background): “Similarly, hackers can successfully hide the malicious code within specially crafted network packets that can evade an intrusion detection system (IDS) without compromising the effectiveness of their attack (see, e.g., Fogla et al., 2006).”
Robust Learning under Uncertain Test Distributions: Relating Covariate Shift to Model Misspecification (2014, background): “Teo et al. (2008) and Globerson & Roweis (2006) also consider an adversarial scenario under changing test set conditions, but they are concerned with corruption or deletion of features rather than covariate shift.”
Learning with Marginalized Corrupted Features (2013, background): “Teo et al. (2008) generalize the worstcase scenario to obtain invariances to transformations such as image rotations or translations.”
Pattern Recognition Systems under Attack: Design Issues and Research Challenges (2014, methodology): “In particular, different variants of the SVM learning algorithm have been proposed to account for a maximum number of worst-case feature manipulations, including deletion, insertion, and rescaling [37, 68, 32].”
Static prediction games for adversarial learning problems (2012, methodology): “We compare a regular support vector machine (SVM), logistic regression (LR), the SVM with trigonometric loss (SVMT, a variant of the SVM which minimizes (33) for the given training data), the worst-case solution SVM for invariances with feature removal (Invar-SVM, Globerson and Roweis, 2006; Teo et ”
Adversarial Biometric Recognition : A review on biometric system security from the adversarial machine-learning perspective (2015, background): “In the context of biometric systems, secure learning techniques can be exploited to design trainable score-level fusion rules, such as those based on game theory, or on the framework of learning with invariances [14], [25]–[27].”