András A. Benczúr

Domokos M. Kelen, Mihály Petreczky, Péter Kersch et al.

In this work, we examine Asymmetric Shapley Values (ASV), a variant of the popular SHAP additive local explanation method. ASV proposes a way to improve model explanations incorporating known causal relations between variables, and is also considered as a way to test for unfair discrimination in model predictions. Unexplored in previous literature, relaxing symmetry in Shapley values can have counter-intuitive consequences for model explanation. To better understand the method, we first show how local contributions correspond to global contributions of variance reduction. Using variance, we demonstrate multiple cases where ASV yields counter-intuitive attributions, arguably producing incorrect results for root-cause analysis. Second, we identify generalized additive models (GAM) as a restricted class for which ASV exhibits desirable properties. We support our arguments by proving multiple theoretical results about the method. Finally, we demonstrate the use of asymmetric attributions on multiple real-world datasets, comparing the results with and without restricted model families using gradient boosting and deep learning models.

Edith Alice Kovács, Anna Ország, Dániel Pfeifer et al.

In this paper we introduce the so-called Generalized Naive Bayes structure as an extension of the Naive Bayes structure. We give a new greedy algorithm that finds a good fitting Generalized Naive Bayes (GNB) probability distribution. We prove that this fits the data at least as well as the probability distribution determined by the classical Naive Bayes (NB). Then, under a not very restrictive condition, we give a second algorithm for which we can prove that it finds the optimal GNB probability distribution, i.e. best fitting structure in the sense of KL divergence. Both algorithms are constructed to maximize the information content and aim to minimize redundancy. Based on these algorithms, new methods for feature selection are introduced. We discuss the similarities and differences to other related algorithms in terms of structure, methodology, and complexity. Experimental results show, that the algorithms introduced outperform the related algorithms in many cases.

Daniel Racz, Martin Gonzalez, Mihaly Petreczky et al.

One of the main theoretical challenges in learning dynamical systems from data is providing upper bounds on the generalization error, that is, the difference between the expected prediction error and the empirical prediction error measured on some finite sample. In machine learning, a popular class of such bounds are the so-called Probably Approximately Correct (PAC) bounds. In this paper, we derive a PAC bound for stable continuous-time linear parameter-varying (LPV) systems. Our bound depends on the H2 norm of the chosen class of the LPV systems, but does not depend on the time interval for which the signals are considered.

Ferenc Béres, Rita Csoma, Tamás Vilmos Michaletzky et al.

We demonstrate the applicability of network embedding to vaccine skepticism, a controversial topic of long-past history. With the Covid-19 pandemic outbreak at the end of 2019, the topic is more important than ever. Only a year after the first international cases were registered, multiple vaccines were developed and passed clinical testing. Besides the challenges of development, testing, and logistics, another factor that might play a significant role in the fight against the pandemic are people who are hesitant to get vaccinated, or even state that they will refuse any vaccine offered to them. Two groups of people commonly referred to as a) pro-vaxxer, those who support vaccinating people b) vax-skeptic, those who question vaccine efficacy or the need for general vaccination against Covid-19. It is very difficult to tell exactly how many people share each of these views. It is even more difficult to understand all the reasoning why vax-skeptic opinions are getting more popular. In this work, our intention was to develop techniques that are able to efficiently differentiate between pro-vaxxer and vax-skeptic content. After multiple data preprocessing steps, we analyzed the tweet text as well as the structure of user interactions on Twitter. We deployed several node embedding and community detection models that scale well for graphs with millions of edges.

Ferenc Béres, István András Seres, András A. Benczúr et al.

Ethereum is the largest public blockchain by usage. It applies an account-based model, which is inferior to Bitcoin's unspent transaction output model from a privacy perspective. Due to its privacy shortcomings, recently several privacy-enhancing overlays have been deployed on Ethereum, such as non-custodial, trustless coin mixers and confidential transactions. In our privacy analysis of Ethereum's account-based model, we describe several patterns that characterize only a limited set of users and successfully apply these quasi-identifiers in address deanonymization tasks. Using Ethereum Name Service identifiers as ground truth information, we quantitatively compare algorithms in recent branch of machine learning, the so-called graph representation learning, as well as time-of-day activity and transaction fee based user profiling techniques. As an application, we rigorously assess the privacy guarantees of the Tornado Cash coin mixer by discovering strong heuristics to link the mixing parties. To the best of our knowledge, we are the first to propose and implement Ethereum user profiling techniques based on quasi-identifiers. Finally, we describe a malicious value-fingerprinting attack, a variant of the Danaan-gift attack, applicable for the confidential transaction overlays on Ethereum. By incorporating user activity statistics from our data set, we estimate the success probability of such an attack.

Bálint Daróczy, Rita Aleksziev, András Benczúr

Hierarchical neural networks are exponentially more efficient than their corresponding "shallow" counterpart with the same expressive power, but involve huge number of parameters and require tedious amounts of training. By approximating the tangent subspace, we suggest a sparse representation that enables switching to shallow networks, GradNet after a very early training stage. Our experiments show that the proposed approximation of the metric improves and sometimes even surpasses the achievable performance of the original network significantly even after a few epochs of training the original feedforward network.

Ferenc Beres, Istvan Andras Seres, Andras A. Benczur

Lightning Network (LN) is designed to amend the scalability and privacy issues of Bitcoin. It's a payment channel network where Bitcoin transactions are issued off chain, onion routed through a private payment path with the aim to settle transactions in a faster, cheaper, and private manner, as they're not recorded in a costly-to-maintain, slow, and public ledger. In this work, we design a traffic simulator to empirically study LN's transaction fees and privacy provisions. The simulator relies on publicly available data of the network structure and generates transactions under assumptions we attempt to validate based on information spread by certain blog posts of LN node owners. Our findings on the estimated revenue from transaction fees are in line with widespread opinion that participation is economically irrational for the majority of large routing nodes who currently hold the network together. Either traffic or transaction fees must increase by orders of magnitude to make payment routing economically viable. We give worst-case estimates for the potential fee increase by assuming strong price competition among the routers. We estimate how current channel structures and pricing policies respond to a potential increase in traffic, how reduction in locked funds on channels would affect the network, and show examples of nodes who are estimated to operate with economically feasible revenue. Even if transactions are onion routed, strong statistical evidence on payment source and destination can be inferred, as many transaction paths only consist of a single intermediary by the side effect of LN's small-world nature. Based on our simulation experiments, we quantitatively characterize the privacy shortcomings of current LN operation, and propose a method to inject additional hops in routing paths to demonstrate how privacy can be strengthened with very little additional transactional cost.

Bálint Daróczy, Rita Aleksziev, András Benczúr

Hierarchical neural networks are exponentially more efficient than their corresponding "shallow" counterpart with the same expressive power, but involve huge number of parameters and require tedious amounts of training. Our main idea is to mathematically understand and describe the hierarchical structure of feedforward neural networks by reparametrization invariant Riemannian metrics. By computing or approximating the tangent subspace, we better utilize the original network via sparse representations that enables switching to shallow networks after a very early training stage. Our experiments show that the proposed approximation of the metric improves and sometimes even surpasses the achievable performance of the original network significantly even after a few epochs of training the original feedforward network.

András A. Benczúr, Levente Kocsis, Róbert Pálovics

The area of online machine learning in big data streams covers algorithms that are (1) distributed and (2) work from data streams with only a limited possibility to store past data. The first requirement mostly concerns software architectures and efficient algorithms. The second one also imposes nontrivial theoretical restrictions on the modeling methods: In the data stream model, older data is no longer available to revise earlier suboptimal modeling decisions as the fresh data arrives. In this article, we provide an overview of distributed software architectures and libraries as well as machine learning models for online learning. We highlight the most important ideas for classification, regression, recommendation, and unsupervised modeling from streaming data, and we show how they are implemented in various distributed data stream processing systems. This article is a reference material and not a survey. We do not attempt to be comprehensive in describing all existing methods and solutions; rather, we give pointers to the most important resources in the field. All related sub-fields, online algorithms, online learning, and distributed data processing are hugely dominant in current research and development with conceptually new research results and software components emerging at the time of writing. In this article, we refer to several survey results, both for distributed data processing and for online machine learning. Compared to past surveys, our article is different because we discuss recommender systems in extended detail.

Bálint Daróczy, Frederick Ayala-Gómez, András Benczúr

Web recommendation services bear great importance in e-commerce, as they aid the user in navigating through the items that are most relevant to her needs. In a typical Web site, long history of previous activities or purchases by the user is rarely available. Hence in most cases, recommenders propose items that are similar to the most recent ones viewed in the current user session. The corresponding task is called session based item-to-item recommendation. For frequent items, it is easy to present item-to-item recommendations by "people who viewed this, also viewed" lists. However, most of the items belong to the long tail, where previous actions are sparsely available. Another difficulty is the so-called cold start problem, when the item has recently appeared and had no time yet to accumulate sufficient number of transactions. In order to recommend a next item in a session in sparse or cold start situations, we also have to incorporate item similarity models. In this paper we describe a probabilistic similarity model based on Random Fields to approximate item-to-item transition probabilities. We give a generative model for the item interactions based on arbitrary distance measures over the items including explicit, implicit ratings and external metadata. The model may change in time to fit better recent events and recommend the next item based on the updated Fisher Information. Our new model outperforms both simple similarity baseline methods and recent item-to-item recommenders, under several different performance metrics and publicly available data sets. We reach significant gains in particular for recommending a new item following a rare item.