Endre søk
Link to record
Permanent link

Direct link
Publikasjoner (10 av 18) Visa alla publikasjoner
Mamaghan, A. M., Tigas, P., Johansson, K. H., Gal, Y., Annadani, Y. & Bauer, S. (2024). Challenges and Considerations in the Evaluation of Bayesian Causal Discovery. In: International Conference on Machine Learning, ICML 2024: . Paper presented at 41st International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024 (pp. 23215-23237). ML Research Press
Åpne denne publikasjonen i ny fane eller vindu >>Challenges and Considerations in the Evaluation of Bayesian Causal Discovery
Vise andre…
2024 (engelsk)Inngår i: International Conference on Machine Learning, ICML 2024, ML Research Press , 2024, s. 23215-23237Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Bayesian Causal Discovery (BCD) offers a principled approach to encapsulating this uncertainty. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, evaluating BCD presents challenges due to the nature of its inferred quantity - the posterior distribution. As a result, the research community has proposed various metrics to assess the quality of the approximate posterior. However, there is, to date, no consensus on the most suitable metric(s) for evaluation. In this work, we reexamine this question by dissecting various metrics and understanding their limitations. Through extensive empirical evaluation, we find that many existing metrics fail to exhibit a strong correlation with the quality of approximation to the true posterior, especially in scenarios with low sample sizes where BCD is most desirable. We highlight the suitability (or lack thereof) of these metrics under two distinct factors: the identifiability of the underlying causal model and the quantity of available data. Both factors affect the entropy of the true posterior, indicating that the current metrics are less fitting in settings of higher entropy. Our findings underline the importance of a more nuanced evaluation of new methods by taking into account the nature of the true posterior, as well as guide and motivate the development of new evaluation procedures for this challenge.

sted, utgiver, år, opplag, sider
ML Research Press, 2024
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-353949 (URN)2-s2.0-85203804991 (Scopus ID)
Konferanse
41st International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024
Merknad

QC 20250922

Tilgjengelig fra: 2024-09-25 Laget: 2024-09-25 Sist oppdatert: 2025-09-22bibliografisk kontrollert
Mamaghan, A. M., Papa, S., Johansson, K. H., Bauer, S. & Dittadi, A. (2024). Exploring the effectiveness of object-centric representations in visual question answering: Comparative insights with foundation models.
Åpne denne publikasjonen i ny fane eller vindu >>Exploring the effectiveness of object-centric representations in visual question answering: Comparative insights with foundation models
Vise andre…
2024 (engelsk)Manuskript (preprint) (Annet vitenskapelig)
Abstract [en]

Object-centric (OC) representations, which represent the state of a visual scene by modeling it as a composition of objects, have the potential to be used in various downstream tasks to achieve systematic compositional generalization and facilitate reasoning. However, these claims have not been thoroughly analyzed yet.Recently, foundation models have demonstrated unparalleled capabilities across diverse domains from language to computer vision, marking them as a potential cornerstone of future research for a multitude of computational tasks.In this paper, we conduct an extensive empirical study on representation learning for downstream Visual Question Answering (VQA), which requires an accurate compositional understanding of the scene. We thoroughly investigate the benefits and trade-offs of OC models and alternative approaches including large pre-trained foundation models on both synthetic and real-world data, and demonstrate a viable way to achieve the best of both worlds. The extensiveness of our study, encompassing over 600 downstream VQA models and 15 different types of upstream representations, also provides several additional insights that we believe will be of interest to the community at large.

HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-359638 (URN)10.48550/arXiv.2407.15589 (DOI)
Merknad

The manuscript is accepted at ICLR 2025 conference

QC 20250211

Tilgjengelig fra: 2025-02-06 Laget: 2025-02-06 Sist oppdatert: 2025-09-23bibliografisk kontrollert
Min, Y., Wei, Y., Wang, P., Wang, X., Li, H., Wu, N., . . . Zeng, J. (2024). From Static to Dynamic Structures: Improving Binding Affinity Prediction with Graph-Based Deep Learning. Advanced Science, 11(40), Article ID 2405404.
Åpne denne publikasjonen i ny fane eller vindu >>From Static to Dynamic Structures: Improving Binding Affinity Prediction with Graph-Based Deep Learning
Vise andre…
2024 (engelsk)Inngår i: Advanced Science, E-ISSN 2198-3844, Vol. 11, nr 40, artikkel-id 2405404Artikkel i tidsskrift (Fagfellevurdert) Published
Abstract [en]

Accurate prediction of protein-ligand binding affinities is an essential challenge in structure-based drug design. Despite recent advances in data-driven methods for affinity prediction, their accuracy is still limited, partially because they only take advantage of static crystal structures while the actual binding affinities are generally determined by the thermodynamic ensembles between proteins and ligands. One effective way to approximate such a thermodynamic ensemble is to use molecular dynamics (MD) simulation. Here, an MD dataset containing 3,218 different protein-ligand complexes is curated, and Dynaformer, a graph-based deep learning model is further developed to predict the binding affinities by learning the geometric characteristics of the protein-ligand interactions from the MD trajectories. In silico experiments demonstrated that the model exhibits state-of-the-art scoring and ranking power on the CASF-2016 benchmark dataset, outperforming the methods hitherto reported. Moreover, in a virtual screening on heat shock protein 90 (HSP90) using Dynaformer, 20 candidates are identified and their binding affinities are further experimentally validated. Dynaformer displayed promising results in virtual drug screening, revealing 12 hit compounds (two are in the submicromolar range), including several novel scaffolds. Overall, these results demonstrated that the approach offer a promising avenue for accelerating the early drug discovery process.

sted, utgiver, år, opplag, sider
Wiley, 2024
Emneord
binding affinity, drug discovery, graph transformer, molecular dynamics
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-366357 (URN)10.1002/advs.202405404 (DOI)001327523600001 ()39206846 (PubMedID)2-s2.0-85202628588 (Scopus ID)
Merknad

QC 20250707

Tilgjengelig fra: 2025-07-07 Laget: 2025-07-07 Sist oppdatert: 2025-07-07bibliografisk kontrollert
Xian, R. P., Stimper, V., Zacharias, M., Dendzik, M., Dong, S., Beaulieu, S., . . . Ernstorfer, R. (2023). A machine learning route between band mapping and band structure. Nature Computational Science, 3(1), 101-114
Åpne denne publikasjonen i ny fane eller vindu >>A machine learning route between band mapping and band structure
Vise andre…
2023 (engelsk)Inngår i: Nature Computational Science, E-ISSN 2662-8457, Vol. 3, nr 1, s. 101-114Artikkel i tidsskrift (Fagfellevurdert) Published
Abstract [en]

The electronic band structure and crystal structure are the two complementary identifiers of solid-state materials. Although convenient instruments and reconstruction algorithms have made large, empirical, crystal structure databases possible, extracting the quasiparticle dispersion (closely related to band structure) from photoemission band mapping data is currently limited by the available computational methods. To cope with the growing size and scale of photoemission data, here we develop a pipeline including probabilistic machine learning and the associated data processing, optimization and evaluation methods for band-structure reconstruction, leveraging theoretical calculations. The pipeline reconstructs all 14 valence bands of a semiconductor and shows excellent performance on benchmarks and other materials datasets. The reconstruction uncovers previously inaccessible momentum-space structural information on both global and local scales, while realizing a path towards integration with materials science databases. Our approach illustrates the potential of combining machine learning and domain knowledge for scalable feature extraction in multidimensional data.

sted, utgiver, år, opplag, sider
Springer Nature, 2023
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-330095 (URN)10.1038/s43588-022-00382-2 (DOI)000905841700001 ()2-s2.0-85145213045 (Scopus ID)
Merknad

QC 20230626

Tilgjengelig fra: 2023-06-26 Laget: 2023-06-26 Sist oppdatert: 2023-06-26bibliografisk kontrollert
Gürtler, N., Blaes, S., Kolev, P., Widmaier, F., Wüthrich, M., Bauer, S., . . . Martius, G. (2023). Benchmarking Offline Reinforcement Learning On Real-Robot Hardware. In: 11th International Conference on Learning Representations, ICLR 2023: . Paper presented at 11th International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1 2023 - May 5 2023. International Conference on Learning Representations, ICLR
Åpne denne publikasjonen i ny fane eller vindu >>Benchmarking Offline Reinforcement Learning On Real-Robot Hardware
Vise andre…
2023 (engelsk)Inngår i: 11th International Conference on Learning Representations, ICLR 2023, International Conference on Learning Representations, ICLR , 2023Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Learning policies from previously recorded data is a promising direction for real-world robotics tasks, as online learning is often infeasible. Dexterous manipulation in particular remains an open problem in its general form. The combination of offline reinforcement learning with large diverse datasets, however, has the potential to lead to a breakthrough in this challenging domain analogously to the rapid progress made in supervised learning in recent years. To coordinate the efforts of the research community toward tackling this problem, we propose a benchmark including: i) a large collection of data for offline learning from a dexterous manipulation platform on two tasks, obtained with capable RL agents trained in simulation; ii) the option to execute learned policies on a real-world robotic system and a simulation for efficient debugging. We evaluate prominent open-sourced offline reinforcement learning algorithms on the datasets and provide a reproducible experimental setup for offline reinforcement learning on real systems. Visit https://sites.google.com/view/benchmarking-offline-rl-real for more details.

sted, utgiver, år, opplag, sider
International Conference on Learning Representations, ICLR, 2023
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-351748 (URN)2-s2.0-85199901138 (Scopus ID)
Konferanse
11th International Conference on Learning Representations, ICLR 2023, Kigali, Rwanda, May 1 2023 - May 5 2023
Merknad

QC 20240813

Tilgjengelig fra: 2024-08-13 Laget: 2024-08-13 Sist oppdatert: 2025-02-05bibliografisk kontrollert
Wang, Q., Sanchez, F. R., McCarthy, R., Bulens, D. C., McGuinness, K., O'Connor, N., . . . Redmond, S. J. (2023). Dexterous robotic manipulation using deep reinforcement learning and knowledge transfer for complex sparse reward-based tasks. Expert Systems, 40(6), Article ID e13205.
Åpne denne publikasjonen i ny fane eller vindu >>Dexterous robotic manipulation using deep reinforcement learning and knowledge transfer for complex sparse reward-based tasks
Vise andre…
2023 (engelsk)Inngår i: Expert Systems, ISSN 0266-4720, E-ISSN 1468-0394, Vol. 40, nr 6, artikkel-id e13205Artikkel i tidsskrift (Fagfellevurdert) Published
Abstract [en]

This paper describes a deep reinforcement learning (DRL) approach that won Phase 1 of the Real Robot Challenge (RRC) 2021, and then extends this method to a more difficult manipulation task. The RRC consisted of using a TriFinger robot to manipulate a cube along a specified positional trajectory, but with no requirement for the cube to have any specific orientation. We used a relatively simple reward function, a combination of a goal-based sparse reward and a distance reward, in conjunction with Hindsight Experience Replay (HER) to guide the learning of the DRL agent (Deep Deterministic Policy Gradient [DDPG]). Our approach allowed our agents to acquire dexterous robotic manipulation strategies in simulation. These strategies were then deployed on the real robot and outperformed all other competition submissions, including those using more traditional robotic control techniques, in the final evaluation stage of the RRC. Here we extend this method, by modifying the task of Phase 1 of the RRC to require the robot to maintain the cube in a particular orientation, while the cube is moved along the required positional trajectory. The requirement to also orient the cube makes the agent less able to learn the task through blind exploration due to increased problem complexity. To circumvent this issue, we make novel use of a Knowledge Transfer (KT) technique that allows the strategies learned by the agent in the original task (which was agnostic to cube orientation) to be transferred to this task (where orientation matters). KT allowed the agent to learn and perform the extended task in the simulator, which improved the average positional deviation from 0.134 to 0.02 m, and average orientation deviation from 142° to 76° during evaluation. This KT concept shows good generalization properties and could be applied to any actor-critic learning algorithm.

sted, utgiver, år, opplag, sider
Wiley, 2023
Emneord
deep reinforcement learning, Real Robot Challenge, robotic manipulation, sim-to-real transfer, transfer reinforcement learning
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-349562 (URN)10.1111/exsy.13205 (DOI)000910515800001 ()2-s2.0-85143421268 (Scopus ID)
Merknad

QC 20240702

Tilgjengelig fra: 2024-07-02 Laget: 2024-07-02 Sist oppdatert: 2025-12-02bibliografisk kontrollert
Mehrjou, A., Soleymani, A., Abyaneh, A., Bhatt, S., Schoelkopf, B. & Bauer, S. (2023). Pyfectious: An individual-level simulator to discover optimal containment policies for epidemic diseases. PloS Computational Biology, 19(1), e1010799, Article ID e1010799.
Åpne denne publikasjonen i ny fane eller vindu >>Pyfectious: An individual-level simulator to discover optimal containment policies for epidemic diseases
Vise andre…
2023 (engelsk)Inngår i: PloS Computational Biology, ISSN 1553-734X, E-ISSN 1553-7358, Vol. 19, nr 1, s. e1010799-, artikkel-id e1010799Artikkel i tidsskrift (Fagfellevurdert) Published
Abstract [en]

Author summaryPyfectious is an agent-based simulator with the capability to serve as an environment for reinforcement learning agents to discover novel control high-resolution agent-based policies that are hard for humans to discover. Pyfectious introduces several novelties which are unprecedented in the existing popular simulators in epidemiology. It constructs the population structure of a city without needing too detailed information by a novel probabilistic assignment method that is unparalleled to existing population synthesizers. The proposed disease propagation algorithm offers a multi-resolution functionality that allows running Pyfectious for large-population cities on normal computers. The modeling details can be easily traded-off with computational demand requiring minimal effort by the end user. The control and monitoring components are designed in an event-triggered fully flexible way by providing a rich action space from which effective policies are hoped to be discovered by advanced RL methods which are otherwise impossible for humans to find due to the immense complexity of the problem. An extensive set of experiments are included to illustrate various aspects of Pyfectious and also to briefly showcase its use as an RL environment which is hoped to help the automatic discovery of epidemic control policies upon bringing together RL scientists and epidemiologists. Simulating the spread of infectious diseases in human communities is critical for predicting the trajectory of an epidemic and verifying various policies to control the devastating impacts of the outbreak. Many existing simulators are based on compartment models that divide people into a few subsets and simulate the dynamics among those subsets using hypothesized differential equations. However, these models lack the requisite granularity to study the effect of intelligent policies that influence every individual in a particular way. In this work, we introduce a simulator software capable of modeling a population structure and controlling the disease's propagation at an individualistic level. In order to estimate the confidence of the conclusions drawn from the simulator, we employ a comprehensive probabilistic approach where the entire population is constructed as a hierarchical random variable. This approach makes the inferred conclusions more robust against sampling artifacts and gives confidence bounds for decisions based on the simulation results. To showcase potential applications, the simulator parameters are set based on the formal statistics of the COVID-19 pandemic, and the outcome of a wide range of control measures is investigated. Furthermore, the simulator is used as the environment of a reinforcement learning problem to find the optimal policies to control the pandemic. The obtained experimental results indicate the simulator's adaptability and capacity in making sound predictions and a successful policy derivation example based on real-world data. As an exemplary application, our results show that the proposed policy discovery method can lead to control measures that produce significantly fewer infected individuals in the population and protect the health system against saturation.

sted, utgiver, år, opplag, sider
Public Library of Science (PLoS), 2023
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-324823 (URN)10.1371/journal.pcbi.1010799 (DOI)000937227800001 ()36689461 (PubMedID)2-s2.0-85147020794 (Scopus ID)
Merknad

QC 20230317

Tilgjengelig fra: 2023-03-17 Laget: 2023-03-17 Sist oppdatert: 2023-03-17bibliografisk kontrollert
Caldarelli, E., Wenk, P., Bauer, S. & Krause, A. (2022). Adaptive Gaussian Process Change Point Detection. In: Chaudhuri, K Jegelka, S Song, L Szepesvari, C Niu, G Sabato, S (Ed.), International conference on machine learning, vol 162: . Paper presented at 38th International Conference on Machine Learning (ICML), JUL 17-23, 2022, Baltimore, MD (pp. 2542-2571). ML Research Press
Åpne denne publikasjonen i ny fane eller vindu >>Adaptive Gaussian Process Change Point Detection
2022 (engelsk)Inngår i: International conference on machine learning, vol 162 / [ed] Chaudhuri, K Jegelka, S Song, L Szepesvari, C Niu, G Sabato, S, ML Research Press , 2022, s. 2542-2571Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

Detecting change points in time series, i.e., points in time at which some observed process suddenly changes, is a fundamental task that arises in many real-world applications, with consequences for safety and reliability. In this work, we propose ADAGA, a novel Gaussian process-based solution to this problem, that leverages a powerful heuristics we developed based on statistical hypothesis testing. In contrast to prior approaches, ADAGA adapts to changes both in mean and covariance structure of the temporal process. In extensive experiments, we show its versatility and applicability to different classes of change points, demonstrating that it is significantly more accurate than current state-of-the-art alternatives.

sted, utgiver, år, opplag, sider
ML Research Press, 2022
Serie
Proceedings of Machine Learning Research, ISSN 2640-3498 ; 162
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-324821 (URN)000899944902026 ()2-s2.0-85163085783 (Scopus ID)
Konferanse
38th International Conference on Machine Learning (ICML), JUL 17-23, 2022, Baltimore, MD
Merknad

QC 20230317

Tilgjengelig fra: 2023-03-17 Laget: 2023-03-17 Sist oppdatert: 2023-07-13bibliografisk kontrollert
Deleu, T., Góis, A., Emezue, C., Rankawat, M., Lacoste-Julien, S., Bauer, S. & Bengio, Y. (2022). Bayesian Structure Learning with Generative Flow Networks. In: Proceedings of the 38th Conference on Uncertainty in Artificial Intelligence, UAI 2022: . Paper presented at 38th Conference on Uncertainty in Artificial Intelligence, UAI 2022, 1 August - 5 August 2022, Eindhoven, Netherlands (pp. 518-528). Association For Uncertainty in Artificial Intelligence (AUAI)
Åpne denne publikasjonen i ny fane eller vindu >>Bayesian Structure Learning with Generative Flow Networks
Vise andre…
2022 (engelsk)Inngår i: Proceedings of the 38th Conference on Uncertainty in Artificial Intelligence, UAI 2022, Association For Uncertainty in Artificial Intelligence (AUAI) , 2022, s. 518-528Konferansepaper, Publicerat paper (Fagfellevurdert)
Abstract [en]

In Bayesian structure learning, we are interested in inferring a distribution over the directed acyclic graph (DAG) structure of Bayesian networks, from data. Defining such a distribution is very challenging, due to the combinatorially large sample space, and approximations based on MCMC are often required. Recently, a novel class of probabilistic models, called Generative Flow Networks (GFlowNets), have been introduced as a general framework for generative modeling of discrete and composite objects, such as graphs. In this work, we propose to use a GFlowNet as an alternative to MCMC for approximating the posterior distribution over the structure of Bayesian networks, given a dataset of observations. Generating a sample DAG from this approximate distribution is viewed as a sequential decision problem, where the graph is constructed one edge at a time, based on learned transition probabilities. Through evaluation on both simulated and real data, we show that our approach, called DAG-GFlowNet, provides an accurate approximation of the posterior over DAGs, and it compares favorably against other methods based on MCMC or variational inference. 

sted, utgiver, år, opplag, sider
Association For Uncertainty in Artificial Intelligence (AUAI), 2022
Emneord
Artificial intelligence, Flow graphs, Probability distributions, Acyclic graphs, Bayesia n networks, Bayesian structure learning, Composite objects, Discrete objects, Flow network, Generative model, Graph structures, Probabilistic models, Sample space, Bayesian networks
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-327050 (URN)2-s2.0-85137616447 (Scopus ID)
Konferanse
38th Conference on Uncertainty in Artificial Intelligence, UAI 2022, 1 August - 5 August 2022, Eindhoven, Netherlands
Merknad

QC 20230523

Tilgjengelig fra: 2023-05-23 Laget: 2023-05-23 Sist oppdatert: 2023-11-13bibliografisk kontrollert
Soleymani, A., Raj, A., Bauer, S., Schölkopf, B. & Besserve, M. (2022). Causal Feature Selection via Orthogonal Search. Transactions on Machine Learning Research, 2022-August
Åpne denne publikasjonen i ny fane eller vindu >>Causal Feature Selection via Orthogonal Search
Vise andre…
2022 (engelsk)Inngår i: Transactions on Machine Learning Research, E-ISSN 2835-8856, Vol. 2022-AugustArtikkel i tidsskrift (Fagfellevurdert) Published
Abstract [en]

The problem of inferring the direct causal parents of a response variable among a large set of explanatory variables is of high practical importance in many disciplines. However, established approaches often scale at least exponentially with the number of explanatory variables, are difficult to extend to nonlinear relationships and are difficult to extend to cyclic data. Inspired by Debiased machine learning methods, we study a one-vs.-the-rest feature selection approach to discover the direct causal parent of the response. We propose an algorithm that works for purely observational data while also offering theoretical guarantees, including the case of partially nonlinear relationships possibly under the presence of cycles. As it requires only one estimation for each variable, our approach is applicable even to large graphs. We demonstrate significant improvements compared to established approaches.

sted, utgiver, år, opplag, sider
Transactions on Machine Learning Research, 2022
HSV kategori
Identifikatorer
urn:nbn:se:kth:diva-361998 (URN)2-s2.0-105000038372 (Scopus ID)
Merknad

QC 20250404

Tilgjengelig fra: 2025-04-03 Laget: 2025-04-03 Sist oppdatert: 2025-04-04bibliografisk kontrollert
Organisasjoner
Identifikatorer
ORCID-id: ORCID iD iconorcid.org/0000-0003-1712-060X