Search

In the natural language processing (NLP) research community, disentangled representation learning hasbecome commonplace in text style transfer and sentiment analysis. Previous studies have demonstrated theutility of extracting style from text corpora in order to augment context-dependent downstream tasks suchas text generation. Within sentiment analysis specifically, disentangled representation learning has beenshown to produce latent representations that can be used to improve downstream classification tasks. In thisstudy, we build upon this existing framework by (1) investigating disentangled representation learning inthe multidimensional task of emotion detection, (2) testing the robustness of this methodology over varyingdatasets, and (3) exploring the interpretability of the produced latent representations. We discover thatclosely following existing disentangled representation learning methods for sentiment analysis in a multi-class setting, performance decreases significantly, and we are unable to effectively distinguish content andstyle in our learned latent representations. Further work is necessary to determine the effectiveness of styledisentanglement for text in multi-class settings using adversarial training.

This paper serves as an investigation in the use of energy-based models for adversarial defense via purification and training. Convergent and non-convergent energy-based models are tasked to remove white-box adversarial signals embedded into images from the CIFAR-10 dataset so that they may be classified correctly. This work presents an analysis behind the stochastic behavior of MCMC sampling for adversarial noise reduction in meta-stable energy basins and the benefits and challenges associated with different regimes of energy-based learning for this task.

This dissertation presents novel approaches and applications of machine learning architectures. In particular, these approaches are based on tools from topological data analysis and are used in conjunction with conventional machine learning methods. Topological data analysis, which is based on algebraic topology, can identify significant global mathematical structures which are out of reach of many other approaches. When we use topology we benefit from generality, and when we use conventional methods we benefit from specificity.This dissertation contains a broad overview of data science and topological data analysis, then transitions to three distinct machine learning applications of these methods. The first application uses linear methods to discover the inherent dimensionality of the manifold given by congressional roll call votes. The second uses persistent homology to identify extremely noisy images in both supervised and unsupervised tasks. The last application uses mapper objects to produce robust classification algorithms. Two additional projects are presented later in the appendix, and are related to the three main applications. The first of these constructs a method to choose optimal optimizers, and the second places mathematical constraints on the structure of renormalization group flows.

As the internet continues to be populated with new devices and emerging technologies, the attack surface grows exponentially. Technology is shifting towards a profit-driven Internet of Things market where security is an afterthought. Traditional defending approaches are no longer sufficient to detect both known and unknown attacks to high accuracy. Machine learning intrusion detection systems have proven their success in identifying unknown attacks with high precision. Nevertheless, machine learning models are also vulnerable to attacks. Adversarial examples can be used to evaluate the robustness of a designed model before it is deployed. Further, using adversarial examples is critical to creating a robust model designed for an adversarial environment. Our work evaluates both traditional machine learning and deep learning models' robustness using the Bot-IoT dataset. Our methodology included two main approaches. First, label poisoning, used to cause incorrect classification by the model. Second, the fast gradient sign method, used to evade detection measures. The experiments demonstrated that an attacker could manipulate or circumvent detection with significant probability., Comment: MDPI Mach. Learn. Knowl. Extr. 2021, 3(2), 333-356; https://www.mdpi.com/2624-800X/1/2/14

Fair classification has become an important topic in machine learning research. While most bias mitigation strategies focus on neural networks, we noticed a lack of work on fair classifiers based on decision trees even though they have proven very efficient. In an up-to-date comparison of state-of-the-art classification algorithms in tabular data, tree boosting outperforms deep learning (Zhang et al. in Expert Syst Appl 82:128–150, 2017). For this reason, we have developed a novel approach of adversarial gradient tree boosting. The objective of the algorithm is to predict the output Y with gradient tree boosting while minimizing the ability of an adversarial neural network to predict the sensitive attribute S. The approach incorporates at each iteration the gradient of the neural network directly in the gradient tree boosting. We empirically assess our approach on four popular data sets and compare against state-of-the-art algorithms. The results show that our algorithm achieves a higher accuracy while obtaining the same level of fairness, as measured using a set of different common fairness definitions.

We propose an online algorithm for sequential learning in the contextual multiarmed bandit setting. Our approach is to partition the context space and, then, optimally combine all of the possible mappings between the partition regions and the set of bandit arms in a data-driven manner. We show that in our approach, the best mapping is able to approximate the best arm selection policy to any desired degree under mild Lipschitz conditions. Therefore, we design our algorithm based on the optimal adaptive combination and asymptotically achieve the performance of the best mapping as well as the best arm selection policy. This optimality is also guaranteed to hold even in adversarial environments since we do not rely on any statistical assumptions regarding the contexts or the loss of the bandit arms. Moreover, we design an efficient implementation for our algorithm using various hierarchical partitioning structures, such as lexicographical or arbitrary position splitting and binary trees (BTs) (and several other partitioning examples). For instance, in the case of BT partitioning, the computational complexity is only log-linear in the number of regions in the finest partition. In conclusion, we provide significant performance improvements by introducing upper bounds (with respect to the best arm selection policy) that are mathematically proven to vanish in the average loss per round sense at a faster rate compared to the state of the art. Our experimental work extensively covers various scenarios ranging from bandit settings to multiclass classification with real and synthetic data. In these experiments, we show that our algorithm is highly superior to the state-of-the-art techniques while maintaining the introduced mathematical guarantees and a computationally decent scalability. IEEE

In this paper, we compare different audio signal representations, including the raw audio waveform and a variety of time-frequency representations, for the task of audio synthesis with Generative Adversarial Networks (GANs). We conduct the experiments on a subset of the NSynth dataset. The architecture follows the benchmark Progressive Growing Wasserstein GAN. We perform experiments both in a fully non-conditional manner as well as conditioning the network on the pitch information. We quantitatively evaluate the generated material utilizing standard metrics for assessing generative models, and compare training and sampling times. We show that complex-valued as well as the magnitude and Instantaneous Frequency of the Short-Time Fourier Transform achieve the best results, and yield fast generation and inversion times. The code for feature extraction, training and evaluating the model is available online., 5 pages, 1 figure, 5 tables, to be published in European Signal Processing Conference (EUSIPCO)

Artificial Intelligence (AI) is one of the hottest topics in our society, especially when it comes to solving data-analysis problems. Industry are conducting their digital shifts, and AI is becoming a cornerstone technology for making decisions out of the huge amount of (sensors-based) data available in the production floor. However, such technology may be disappointing when deployed in real conditions. Despite good theoretical performances and high accuracy when trained and tested in isolation, a Machine-Learning (M-L) model may provide degraded performances in real conditions. One reason may be fragility in treating properly unexpected or perturbed data. The objective of the paper is therefore to study the robustness of seven M-L and Deep-Learning (D-L) algorithms, when classifying univariate time-series under perturbations. A systematic approach is proposed for artificially injecting perturbations in the data and for evaluating the robustness of the models. This approach focuses on two perturbations that are likely to happen during data collection. Our experimental study, conducted on twenty sensors’ datasets from the public University of California Riverside (UCR) repository, shows a great disparity of the models’ robustness under data quality degradation. Those results are used to analyse whether the impact of such robustness can be predictable—thanks to decision trees—which would prevent us from testing all perturbations scenarios. Our study shows that building such a predictor is not straightforward and suggests that such a systematic approach needs to be used for evaluating AI models’ robustness.

Deep unsupervised representation learning has recently led to new approaches in the field of Unsupervised Anomaly Detection (UAD) in brain MRI. The main principle behind these works is to learn a model of normal anatomy by learning to compress and recover healthy data. This allows to spot abnormal structures from erroneous recoveries of compressed, potentially anomalous samples. The concept is of great interest to the medical image analysis community as it i) relieves from the need of vast amounts of manually segmented training data—a necessity for and pitfall of current supervised Deep Learning—and ii) theoretically allows to detect arbitrary, even rare pathologies which supervised approaches might fail to find. To date, the experimental design of most works hinders a valid comparison, because i) they are evaluated against different datasets and different pathologies, ii) use different image resolutions and iii) different model architectures with varying complexity. The intent of this work is to establish comparability among recent methods by utilizing a single architecture, a single resolution and the same dataset(s). Besides providing a ranking of the methods, we also try to answer questions like i) how many healthy training subjects are needed to model normality and ii) if the reviewed approaches are also sensitive to domain shift. Further, we identify open challenges and provide suggestions for future community efforts and research directions.

Agents that perform intelligent tasks interacting with humans in a seamless manner are becoming a reality. In contexts in which interactions among agents repeat over time, they might evolve from a cooperative to a competitive attitude, and vice versa, depending on environmental factors and other contextual circumstances. We provide a framework to model transitions between competition and cooperation in a community of agents. Competition is dealt with through the paradigm of adversarial risk analysis, which provides a disagreement solution; implicitly, we minimize the distance to such solution. Cooperation is handled through a concept of maximal separation from the disagreement solution. Mixtures of both problems are used to refer to in-between behaviour. We illustrate the ideas with several simulations in relation with a group of robots. Our motivation is the constitution of communities of robotic agents that interact among them and with one or more users., This research is supported by Singapore Ministry of Education Academic Research Fund Tier 2, MOE2016-T2-2-156.

Many real-world decision-making problems involve multiple conflicting objectives that can not be optimized simultaneously without a compromise. Such problems are known as multi-objective Markov decision processes and they constitute a significant challenge for conventional single-objective reinforcement learning methods, especially when an optimal compromise cannot be determined beforehand. Multi-objective reinforcement learning methods address this challenge by finding an optimal coverage set of non-dominated policies that can satisfy any user's preference in solving the problem. However, this is achieved with costs of computational complexity, time consumption, and lack of adaptability to non-stationary environment dynamics. In order to address these limitations, there is a need for adaptive methods that can solve the problem in an online and robust manner. In this paper, we propose a novel developmental method that utilizes the adversarial self-play between an intrinsically motivated preference exploration component, and a policy coverage set optimization component that robustly evolves a convex coverage set of policies to solve the problem using preferences proposed by the former component. We show experimentally the effectiveness of the proposed method in comparison to state-of-the-art multi-objective reinforcement learning methods in stationary and non-stationary environments.

Classification techniques are widely used in security settings in which data can be deliberately manipulated by an adversary trying to evade detection and achieve some benefit. However, traditional classification systems are not robust to such data modifications. Most attempts to enhance classification algorithms in adversarial environments have focused on game theoretical ideas under strong underlying common knowledge assumptions, which are not actually realistic in security domains. We provide an alternative framework to such problems based on adversarial risk analysis which we illustrate with examples. Computational, implementation and robustness issues are discussed., R.N. acknowledges support the Spanish Ministry for his grant FPU15-03636. The work of D.R.I. is supported by the Spanish Ministry program MTM2017-86875-C3-1-R and the AXA-ICMAT Chair on Adversarial Risk Analysis. This work has also been partially supported by the Spanish Ministry of Economy through the Severo Ochoa Program for Centers of Excellence in R&D (SEV-2015-0554), the project MTM2015-72907-EXP and the EU's Horizon 2020 project 740920 CYBECO (Supporting Cyberinsurance from a Behavioural Choice Perspective). F.R. acknowledges the contribution of the Comunidad de Madrid through its Chair of Excellence programme. We are grateful for the suggestions of the referees.

It is of significant importance for any classification and recognition system, which claims near or better than human performance to be immune to small perturbations in the dataset. Researchers found out that neural networks are not very robust to small perturbations and can easily be fooled to persistently misclassify by adding a particular class of noise in the test data. This, so-called adversarial noise severely deteriorates the performance of neural networks, which otherwise perform really well on unperturbed dataset. It has been recently proposed that neural networks can be made robust against adversarial noise by training them using the data corrupted with adversarial noise itself. Following this approach, in this paper, we propose a new mechanism to generate a powerful adversarial noise model based on K-support norm to train neural networks. We tested our approach on two benchmark datasets, namely the MNIST and STL-10, using muti-layer perceptron and convolutional neural networks. Experimental results demonstrate that neural networks trained with the proposed technique show significant improvement in robustness as compared to state-of-the-art techniques.