Change search
Link to record
Permanent link

Direct link
Publications (10 of 51) Show all publications
Wang, Z., Samsten, I., Miliou, I., Mochaourab, R. & Papapetrou, P. (2024). Glacier: guided locally constrained counterfactual explanations for time series classification. Machine Learning
Open this publication in new window or tab >>Glacier: guided locally constrained counterfactual explanations for time series classification
Show others...
2024 (English)In: Machine Learning, ISSN 0885-6125, E-ISSN 1573-0565Article in journal (Refereed) Epub ahead of print
Abstract [en]

In machine learning applications, there is a need to obtain predictive models of high performance and, most importantly, to allow end-users and practitioners to understand and act on their predictions. One way to obtain such understanding is via counterfactuals, that provide sample-based explanations in the form of recommendations on which features need to be modified from a test example so that the classification outcome of a given classifier changes from an undesired outcome to a desired one. This paper focuses on the domain of time series classification, more specifically, on defining counterfactual explanations for univariate time series. We propose Glacier, a model-agnostic method for generating locally-constrained counterfactual explanations for time series classification using gradient search either on the original space or on a latent space that is learned through an auto-encoder. An additional flexibility of our method is the inclusion of constraints on the counterfactual generation process that favour applying changes to particular time series points or segments while discouraging changing others. The main purpose of these constraints is to ensure more reliable counterfactuals, while increasing the efficiency of the counterfactual generation process. Two particular types of constraints are considered, i.e., example-specific constraints and global constraints. We conduct extensive experiments on 40 datasets from the UCR archive, comparing different instantiations of Glacier against three competitors. Our findings suggest that Glacier outperforms the three competitors in terms of two common metrics for counterfactuals, i.e., proximity and compactness. Moreover, Glacier obtains comparable counterfactual validity compared to the best of the three competitors. Finally, when comparing the unconstrained variant of Glacier to the constraint-based variants, we conclude that the inclusion of example-specific and global constraints yields a good performance while demonstrating the trade-off between the different metrics. © The Author(s) 2024.

Place, publisher, year, edition, pages
Springer, 2024
Keywords
Economic and social effects; Learning systems; Time series; Counterfactual explanation; Counterfactuals; Deep learning; Generation process; Global constraints; Interpretability; Machine learning applications; Performance; Predictive models; Time series classifications; Deep learning
National Category
Computer and Information Sciences
Identifiers
urn:nbn:se:ri:diva-72856 (URN)10.1007/s10994-023-06502-x (DOI)2-s2.0-85187677577 (Scopus ID)
Note

This work was funded in part by the Digital Futures cross-disciplinary research centre in Sweden, and the EXTREMUM collaborative project ( https://datascience.dsv.su.se/projects/extremum.html ).

Available from: 2024-05-14 Created: 2024-05-14 Last updated: 2024-05-23Bibliographically approved
Mochaourab, R., Sinha, S., Greenstein, S. & Papapetrou, P. (2023). Demonstrator on Counterfactual Explanations for Differentially Private Support Vector Machines. In: Lecture Notes in Computer Science : . Paper presented at 22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022. Grenoble. 19 September 2022 through 23 September 2022 (pp. 662-666). Springer Science and Business Media Deutschland GmbH, 13718
Open this publication in new window or tab >>Demonstrator on Counterfactual Explanations for Differentially Private Support Vector Machines
2023 (English)In: Lecture Notes in Computer Science , Springer Science and Business Media Deutschland GmbH , 2023, Vol. 13718, p. 662-666Conference paper, Published paper (Refereed)
Abstract [en]

We demonstrate the construction of robust counterfactual explanations for support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. Privacy preservation is essential when dealing with sensitive data, such as in applications within the health domain. In addition, providing explanations for machine learning predictions is an important requirement within so-called high risk applications, as referred to in the EU AI Act. Thus, the innovative aspects of this work correspond to studying the interaction between three desired aspects: accuracy, privacy, and explainability. The SVM classification accuracy is affected by the privacy mechanism through the introduced perturbations in the classifier weights. Consequently, we need to consider a trade-off between accuracy and privacy. In addition, counterfactual explanations, which quantify the smallest changes to selected data instances in order to change their classification, may become not credible when we have data privacy guarantees. Hence, robustness for counterfactual explanations is needed in order to create confidence about the credibility of the explanations. Our demonstrator provides an interactive environment to show the interplay between the considered aspects of accuracy, privacy, and explainability. 

Place, publisher, year, edition, pages
Springer Science and Business Media Deutschland GmbH, 2023
Keywords
Counterfactual explanations, Differential privacy, Support vector machines, Economic and social effects, Sensitive data, Classification accuracy, Counterfactual explanation, Counterfactuals, Differential privacies, Machine-learning, Privacy mechanisms, Privacy preservation, Sensitive datas, Support vector machine classification, Support vectors machine
National Category
Computer Sciences
Identifiers
urn:nbn:se:ri:diva-65393 (URN)10.1007/978-3-031-26422-1_52 (DOI)2-s2.0-85150995194 (Scopus ID)9783031264214 (ISBN)
Conference
22nd Joint European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2022. Grenoble. 19 September 2022 through 23 September 2022
Note

Correspondence Address: Mochaourab, R.; Digital Systems Division, Sweden; email: rami.mochaourab@ri.se; Funding text 1: Acknowledgements. This work has been supported by the Digital Futures center (https://www.digitalfutures.kth.se) within the project “EXTREMUM: Explainable and Ethical Machine Learning for Knowledge Discovery from Medical Data Sources”.

Available from: 2023-06-15 Created: 2023-06-15 Last updated: 2024-02-12Bibliographically approved
Mochaourab, R., Sinha, S., Greenstein, S. & Papapetrou, P. (2022). Demonstrator on Counterfactual Explanations for Differentially Private Support Vector Machines. In: Proc of ECML PKDD 2022: . Paper presented at European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2022), Demo track. Sep. 19-23, 2022..
Open this publication in new window or tab >>Demonstrator on Counterfactual Explanations for Differentially Private Support Vector Machines
2022 (English)In: Proc of ECML PKDD 2022, 2022Conference paper, Published paper (Refereed)
Abstract [en]

We demonstrate the construction of robust counterfactual explanations for support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. Privacy preservation is essential when dealing with sensitive data, such as in applications within the health domain. In addition, providing explanations for machine learning predictions is an important requirement within so-called high risk applications, as referred to in the EU AI Act. Thus, the innovative aspects of this work correspond to studying the interaction between three desired aspects: accuracy, privacy, and explainability. The SVM classification accuracy is affected by the privacy mechanism through the introduced perturbations in the classifier weights. Consequently, we need to consider a trade-off between accuracy and privacy. In addition, counterfactual explanations, which quantify the smallest changes to selected data instances in order to change their classification, may become not credible when we have data privacy guarantees. Hence, robustness for counterfactual explanations is needed in order to create confidence about the credibility of the explanations. Our demonstrator provides an interactive environment to show the interplay between the considered aspects of accuracy, privacy, and explainability.

Keywords
Counterfactual Explanations · Support Vector Machines · Differential Privacy.
National Category
Computer Sciences
Identifiers
urn:nbn:se:ri:diva-60835 (URN)
Conference
European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2022), Demo track. Sep. 19-23, 2022.
Available from: 2022-10-17 Created: 2022-10-17 Last updated: 2024-02-12Bibliographically approved
Mochaourab, R., Venkitaraman, A., Samsten, I., Papapetrou, P. & Rojas, C. (2022). Post-hoc Explainability for Time Series Classification: Towards a Signal Processing Perspective. IEEE signal processing magazine (Print), 39(4), 119-129
Open this publication in new window or tab >>Post-hoc Explainability for Time Series Classification: Towards a Signal Processing Perspective
Show others...
2022 (English)In: IEEE signal processing magazine (Print), ISSN 1053-5888, E-ISSN 1558-0792, Vol. 39, no 4, p. 119-129Article in journal (Refereed) Published
Abstract [en]

Time series data correspond to observations of phenomena that are recorded over time [1]. Such data are encountered regularly in a wide range of applications, such as speech and music recognition, monitoring health and medical diagnosis, financial analysis, motion tracking, and shape identification, to name a few. With such a diversity of applications and the large variations in their characteristics, time series classification is a complex and challenging task. One of the fundamental steps in the design of time series classifiers is that of defining or constructing the discriminant features that help differentiate between classes. This is typically achieved by designing novel representation techniques [2] that transform the raw time series data to a new data domain, where subsequently a classifier is trained on the transformed data, such as one-nearest neighbors [3] or random forests [4]. In recent time series classification approaches, deep neural network models have been employed that are able to jointly learn a representation of time series and perform classification [5]. In many of these sophisticated approaches, the discriminant features tend to be complicated to analyze and interpret, given the high degree of nonlinearity.

National Category
Signal Processing Computer Sciences
Identifiers
urn:nbn:se:ri:diva-58800 (URN)10.1109/MSP.2022.3155955 (DOI)2-s2.0-85133840717 (Scopus ID)
Available from: 2022-03-09 Created: 2022-03-09 Last updated: 2023-06-07Bibliographically approved
Wang, Z., Samsten, I., Mochaourab, R. & Papapetrou, P. (2021). Learning Time Series Counterfactuals via Latent Space Representations. In: Soares C., Torgo L. (eds) Discovery Science. DS 2021. Lecture Notes in Computer Science, vol 12986. Springer, Cham. https://doi.org/10.1007/978-3-030-88942-5_29: . Paper presented at Discovery Science (pp. 369-384).
Open this publication in new window or tab >>Learning Time Series Counterfactuals via Latent Space Representations
2021 (English)In: Soares C., Torgo L. (eds) Discovery Science. DS 2021. Lecture Notes in Computer Science, vol 12986. Springer, Cham. https://doi.org/10.1007/978-3-030-88942-5_29, 2021, p. 369-384Conference paper, Published paper (Refereed)
National Category
Computer Sciences
Identifiers
urn:nbn:se:ri:diva-58799 (URN)10.1007/978-3-030-88942-5_29 (DOI)
Conference
Discovery Science
Available from: 2022-03-09 Created: 2022-03-09 Last updated: 2023-06-07Bibliographically approved
Mochaourab, R., Sinha, S., Greenstein, S. & Papapetrou, P. (2021). Robust Counterfactual Explanations for Privacy-Preserving SVM. In: Proc of ICML 2021: . Paper presented at International Conference on Machine Learning (ICML 2021), Workshop on Socially Responsible Machine Learning.
Open this publication in new window or tab >>Robust Counterfactual Explanations for Privacy-Preserving SVM
2021 (English)In: Proc of ICML 2021, 2021Conference paper, Published paper (Refereed)
Abstract [en]

We consider counterfactual explanations for privacy-preserving support vector machines (SVM), where the privacy mechanism that publicly releases the classifier guarantees differential privacy. While privacy preservation is essential when dealing with sensitive data, there is a consequent degradation in the classification accuracy due to the introduced perturbations in the classifier weights. Therefore, counterfactual explanations need to be made robust against such perturbations in order to ensure, with high confidence, that the explanations are valid. In this work, we suitably model the uncertainties in the SVM weights and formulate the robust counterfactual explanation problem. Then, we study optimal and efficient suboptimal algorithms for its solution. Experimental results illustrate the connections between privacy levels, classifier accuracy, and the confidence levels that validate the counterfactual explanations.

National Category
Computer Sciences
Identifiers
urn:nbn:se:ri:diva-55465 (URN)
Conference
International Conference on Machine Learning (ICML 2021), Workshop on Socially Responsible Machine Learning
Available from: 2021-07-18 Created: 2021-07-18 Last updated: 2024-02-12Bibliographically approved
Mochaourab, R. & Oechtering, T. J. (2018). Private Filtering for Hidden Markov Models. IEEE Signal Processing Letters, 25(6)
Open this publication in new window or tab >>Private Filtering for Hidden Markov Models
2018 (English)In: IEEE Signal Processing Letters, ISSN 1070-9908, E-ISSN 1558-2361, Vol. 25, no 6Article in journal (Refereed) Published
Abstract [en]

Consider a hidden Markov model describing a system with two types of states: a monitored state and a private state. The two types of states are dependent and evolve jointly according to a Markov process with a stationary transition probability. It is desired to reveal the monitored states to a receiver but hide the private states. For this purpose, a privacy filter is necessary which suitably perturbs the monitored states before communication to the receiver. Our objective is to design the privacy filter to optimize the trade-off between monitoring accuracy and privacy, measured through a time-invariant distortion measure and Shannon's equivocation, respectively. As the optimal privacy filter is difficult to compute using dynamic programming, we adopt a suboptimal greedy approach through which the privacy filter can be computed efficiently. Here, the greedy approach has the additional advantage of not being restricted to finite time horizon setups. Simulations show the superiority of the approach compared to a privacy filter which only adds independent noise to the observations. 

Keywords
Dynamic programming, Greedy algorithm, Hidden Markov models, Privacy, Data privacy, Economic and social effects, Trellis codes, Distortion measures, Finite time horizon, Greedy algorithms, Greedy approaches, Independent noise, Monitoring accuracy, Shannon's equivocation, Transition probabilities
National Category
Natural Sciences
Identifiers
urn:nbn:se:ri:diva-33760 (URN)10.1109/LSP.2018.2827878 (DOI)2-s2.0-85045612606 (Scopus ID)
Available from: 2018-05-07 Created: 2018-05-07 Last updated: 2023-06-07Bibliographically approved
Ok, J., Se-Young, Y., Proutiere, A. & Mochaourab, R. (2017). Collaborative Clustering: Sample Complexity and Efficient Algorithms. In: Proceedings of Machine Learning Research 76:1–42, 2017: (pp. 288-329). , 76
Open this publication in new window or tab >>Collaborative Clustering: Sample Complexity and Efficient Algorithms
2017 (English)In: Proceedings of Machine Learning Research 76:1–42, 2017, 2017, Vol. 76, p. 288-329Conference paper, Published paper (Refereed)
National Category
Probability Theory and Statistics
Identifiers
urn:nbn:se:ri:diva-56403 (URN)
Available from: 2021-09-10 Created: 2021-09-10 Last updated: 2023-06-07Bibliographically approved
Mochaourab, R., Bjornson, E. & Bengtsson, M. (2016). Adaptive Pilot Clustering in Heterogeneous Massive MIMO Networks. IEEE Transactions on Wireless Communications, 15(8), 5555-5568
Open this publication in new window or tab >>Adaptive Pilot Clustering in Heterogeneous Massive MIMO Networks
2016 (English)In: IEEE Transactions on Wireless Communications, Vol. 15, no 8, p. 5555-5568Article in journal (Refereed) Published
Abstract [en]

We consider the uplink of a cellular massive multiple-input multiple-output network. Acquiring channel state information at the base stations (BSs) requires uplink pilot signaling. Since the number of orthogonal pilot sequences is limited by the channel coherence, pilot reuse across cells is necessary to achieve high spectral efficiency. However, finding efficient pilot reuse patterns is non-trivial, especially in practical asymmetric BS deployments. We approach this problem using the coalitional game theory. Each BS has a few unique pilots and can form coalitions with other BSs to gain access to more pilots. The BSs in a coalition, thus, benefit from serving more users in their cells at the expense of higher pilot contamination and interference. Given that a cell’s average spectral efficiency depends on the overall pilot reuse pattern, the suitable coalitional game model is in the partition form. We develop a low-complexity distributed coalition formation based on individual stability. By incorporating a BS intercommunication budget constraint, we are able to control the overhead in message exchange between the BSs and ensure the algorithm’s convergence to a solution of the game called individually stable coalition structure. Simulation results reveal fast algorithmic convergence and substantial performance gains over the baseline schemes with no pilot reuse, full pilot reuse, or random pilot reuse pattern.

Keywords
Massive MIMO networks, spectral efficiency, pilot contamination, coalitional game theory
National Category
Telecommunications
Identifiers
urn:nbn:se:ri:diva-55539 (URN)10.1109/TWC.2016.2561289 (DOI)
Available from: 2021-08-04 Created: 2021-08-04 Last updated: 2023-06-07Bibliographically approved
Brandt, R., Mochaourab, R. & Bengtsson, M. (2016). Distributed Long-Term Base Station Clustering in Cellular Networks using Coalition Formation. IEEE Transactions on Signal and Information Processing over Networks, 2(3), 362-375
Open this publication in new window or tab >>Distributed Long-Term Base Station Clustering in Cellular Networks using Coalition Formation
2016 (English)In: IEEE Transactions on Signal and Information Processing over Networks, Vol. 2, no 3, p. 362-375Article in journal (Refereed) Published
National Category
Communication Systems
Identifiers
urn:nbn:se:ri:diva-56400 (URN)10.1109/TSIPN.2016.2548781 (DOI)
Available from: 2021-09-10 Created: 2021-09-10 Last updated: 2023-06-07Bibliographically approved
Organisations
Identifiers
ORCID iD: ORCID iD iconorcid.org/0000-0003-0995-9835

Search in DiVA

Show all publications