Ponències/Comunicacions de congressoshttp://hdl.handle.net/2117/39742024-03-28T08:48:20Z2024-03-28T08:48:20ZAnalyzing text representations by measuring task alignmentGonzález Gutiérrez, CésarPrimadhanty, AudiCazzaro, FrancescoQuattoni, Ariadna Julietahttp://hdl.handle.net/2117/4054892024-03-28T08:04:24Z2024-03-28T08:02:47ZAnalyzing text representations by measuring task alignment
González Gutiérrez, César; Primadhanty, Audi; Cazzaro, Francesco; Quattoni, Ariadna Julieta
Textual representations based on pre-trained language models are key, especially in few-shot learning scenarios. What makes a representation good for text classification? Is it due to the geometric properties of the space or because it is well aligned with the task? We hypothesize the second claim. To test it, we develop a task alignment score based on hierarchical clustering that measures alignment at different levels of granularity. Our experiments on text classification validate our hypothesis by showing that task alignment can explain the classification performance of a given representation.
2024-03-28T08:02:47ZGonzález Gutiérrez, CésarPrimadhanty, AudiCazzaro, FrancescoQuattoni, Ariadna JulietaTextual representations based on pre-trained language models are key, especially in few-shot learning scenarios. What makes a representation good for text classification? Is it due to the geometric properties of the space or because it is well aligned with the task? We hypothesize the second claim. To test it, we develop a task alignment score based on hierarchical clustering that measures alignment at different levels of granularity. Our experiments on text classification validate our hypothesis by showing that task alignment can explain the classification performance of a given representation.GMX: Instruction set extensions for fast, scalable, and efficient genome sequence alignmentDoblas Font, MaxLostes Cazorla, OscarAguado Puig, QuimCebry, NicholasFontova Muste, PauBatten, ChristopherMarco Sola, SantiagoMoretó Planas, Miquelhttp://hdl.handle.net/2117/4054882024-03-28T08:47:17Z2024-03-28T07:34:55ZGMX: Instruction set extensions for fast, scalable, and efficient genome sequence alignment
Doblas Font, Max; Lostes Cazorla, Oscar; Aguado Puig, Quim; Cebry, Nicholas; Fontova Muste, Pau; Batten, Christopher; Marco Sola, Santiago; Moretó Planas, Miquel
Sequence alignment remains a fundamental problem in computer science with practical applications ranging from pattern matching to computational biology. The ever-increasing volumes of genomic data produced by modern DNA sequencers motivate improved software and hardware sequence alignment accelerators that scale with longer sequence lengths and high error rates without losing accuracy. Furthermore, the wide variety of use cases requiring sequence alignment demands flexible and efficient solutions that can match or even outperform expensive application-specific accelerators.
To address these challenges, we propose GMX, a set of ISA extensions that enable efficient sequence alignment computations based on dynamic programming (DP). GMX extensions provide the basic building-block operations to perform fast tile-wise computations of the DP matrix, reducing the memory footprint and allowing easy integration into widely-used algorithms and tools. Furthermore, we provide an efficient hardware implementation that integrates GMX extensions in a RISC-V-based edge system-on-chip (SoC). Compared to widely-used software implementations, our hardware-software co-design leveraging GMX extensions obtains speed-ups from 25–265 ×, scaling to megabyte-long sequences. Compared to domain-specific accelerators (DSA), we demonstrate that GMX-accelerated implementations demand significantly less memory bandwidth, requiring less area per processing element (PE). As a result, a single GMX-enabled core achieves a throughput per area between 0.35-0.52 × that of state-of-the-art DSAs while being more flexible and reusing the core’s resources. Post-place-and-route results for a GMX-enhanced SoC in 22nm technology shows that GMX extensions only account for 1.7% of the overall area while consuming just 8.47mW. We conclude that GMX extensions represent versatile and scalable ISA additions to improve the performance of genome analysis tools and other use cases that require fast and efficient sequence alignment.
2024-03-28T07:34:55ZDoblas Font, MaxLostes Cazorla, OscarAguado Puig, QuimCebry, NicholasFontova Muste, PauBatten, ChristopherMarco Sola, SantiagoMoretó Planas, MiquelSequence alignment remains a fundamental problem in computer science with practical applications ranging from pattern matching to computational biology. The ever-increasing volumes of genomic data produced by modern DNA sequencers motivate improved software and hardware sequence alignment accelerators that scale with longer sequence lengths and high error rates without losing accuracy. Furthermore, the wide variety of use cases requiring sequence alignment demands flexible and efficient solutions that can match or even outperform expensive application-specific accelerators.
To address these challenges, we propose GMX, a set of ISA extensions that enable efficient sequence alignment computations based on dynamic programming (DP). GMX extensions provide the basic building-block operations to perform fast tile-wise computations of the DP matrix, reducing the memory footprint and allowing easy integration into widely-used algorithms and tools. Furthermore, we provide an efficient hardware implementation that integrates GMX extensions in a RISC-V-based edge system-on-chip (SoC). Compared to widely-used software implementations, our hardware-software co-design leveraging GMX extensions obtains speed-ups from 25–265 ×, scaling to megabyte-long sequences. Compared to domain-specific accelerators (DSA), we demonstrate that GMX-accelerated implementations demand significantly less memory bandwidth, requiring less area per processing element (PE). As a result, a single GMX-enabled core achieves a throughput per area between 0.35-0.52 × that of state-of-the-art DSAs while being more flexible and reusing the core’s resources. Post-place-and-route results for a GMX-enhanced SoC in 22nm technology shows that GMX extensions only account for 1.7% of the overall area while consuming just 8.47mW. We conclude that GMX extensions represent versatile and scalable ISA additions to improve the performance of genome analysis tools and other use cases that require fast and efficient sequence alignment.Introducing social robots to assess frailty in older adultsCivit Bertran, AniolAndriella, AntonioBarrué Subirana, CristianAntonio, MaiteBoqué, ConcepciónAlenyà Ribas, Guillemhttp://hdl.handle.net/2117/4051692024-03-22T13:10:14Z2024-03-22T13:03:01ZIntroducing social robots to assess frailty in older adults
Civit Bertran, Aniol; Andriella, Antonio; Barrué Subirana, Cristian; Antonio, Maite; Boqué, Concepción; Alenyà Ribas, Guillem
Frailty is a crucial indicator in determining the well-being of older adults in terms of their health. With the growing number of elderly people, the demand for geriatricians is increasing, which means that they have less time to spend with each patient. The current methods for frailty assessment use simple tests that are time-consuming and do not require specific medical expertise. To address this issue, this paper proposes the use of social robots to assess frailty autonomously. It presents a practical proposal that defines the robot’s behavior and explains the design and implementation concepts. Finally, it discusses some of the challenges that may arise from introducing social robots as frailty evaluators.
2024-03-22T13:03:01ZCivit Bertran, AniolAndriella, AntonioBarrué Subirana, CristianAntonio, MaiteBoqué, ConcepciónAlenyà Ribas, GuillemFrailty is a crucial indicator in determining the well-being of older adults in terms of their health. With the growing number of elderly people, the demand for geriatricians is increasing, which means that they have less time to spend with each patient. The current methods for frailty assessment use simple tests that are time-consuming and do not require specific medical expertise. To address this issue, this paper proposes the use of social robots to assess frailty autonomously. It presents a practical proposal that defines the robot’s behavior and explains the design and implementation concepts. Finally, it discusses some of the challenges that may arise from introducing social robots as frailty evaluators.The K-Robinson Foulds measures for labeled treesKhayatian, ElaheValiente Feruglio, Gabriel AlejandroZhang, Louxinhttp://hdl.handle.net/2117/4050522024-03-25T01:41:44Z2024-03-21T07:49:22ZThe K-Robinson Foulds measures for labeled trees
Khayatian, Elahe; Valiente Feruglio, Gabriel Alejandro; Zhang, Louxin
Investigating the mutational history of tumor cells is important for understanding the underlying mechanisms of cancer and its evolution. Now that the evolution of tumor cells is modeled using labeled trees, researchers are motivated to propose different measures for the comparison of mutation trees and other labeled trees. While the Robinson-Foulds distance is widely used for the comparison of phylogenetic trees, it has weaknesses when it is applied to labeled trees. Here, k-Robinson-Foulds dissimilarity measures are introduced for labeled tree comparison.
2024-03-21T07:49:22ZKhayatian, ElaheValiente Feruglio, Gabriel AlejandroZhang, LouxinInvestigating the mutational history of tumor cells is important for understanding the underlying mechanisms of cancer and its evolution. Now that the evolution of tumor cells is modeled using labeled trees, researchers are motivated to propose different measures for the comparison of mutation trees and other labeled trees. While the Robinson-Foulds distance is widely used for the comparison of phylogenetic trees, it has weaknesses when it is applied to labeled trees. Here, k-Robinson-Foulds dissimilarity measures are introduced for labeled tree comparison.Evolution of the participation of women in university computer science studies in Spain and in EuropeNebot Castells, M. ÀngelaMúgica Álvarez, Franciscohttp://hdl.handle.net/2117/4050512024-03-25T01:35:22Z2024-03-21T07:23:22ZEvolution of the participation of women in university computer science studies in Spain and in Europe
Nebot Castells, M. Àngela; Múgica Álvarez, Francisco
This paper presents a comprehensive examination of the current state of underrepresentation of women in computer science (CS) degrees in Spain. It offers an overview of the situation across autonomous communities in Spain, highlighting the evolving trends within the country, regional variations, and the distinctions between public and private universities. The research employs descriptive statistical methods to analyze data obtained from the Spanish Ministry of Universities, Eurostat, as well as recent literature reviews. Furthermore, the study contextualizes the evolution of female enrollments in Spain by comparing it with data from other European countries. The primary objective is to gain a thorough understanding of the broader Spanish landscape and suggest future research directions aimed at reducing the gender gap in CS studies.
2024-03-21T07:23:22ZNebot Castells, M. ÀngelaMúgica Álvarez, FranciscoThis paper presents a comprehensive examination of the current state of underrepresentation of women in computer science (CS) degrees in Spain. It offers an overview of the situation across autonomous communities in Spain, highlighting the evolving trends within the country, regional variations, and the distinctions between public and private universities. The research employs descriptive statistical methods to analyze data obtained from the Spanish Ministry of Universities, Eurostat, as well as recent literature reviews. Furthermore, the study contextualizes the evolution of female enrollments in Spain by comparing it with data from other European countries. The primary objective is to gain a thorough understanding of the broader Spanish landscape and suggest future research directions aimed at reducing the gender gap in CS studies.Applying generative models and transfer learning to physiological data classificationNúñez Rodríguez, José FernandoArjona Martínez, JamieTormos Llorente, AdriánGarcia Gasulla, DarioBéjar Alonso, Javierhttp://hdl.handle.net/2117/4049852024-03-24T08:12:45Z2024-03-20T06:46:04ZApplying generative models and transfer learning to physiological data classification
Núñez Rodríguez, José Fernando; Arjona Martínez, Jamie; Tormos Llorente, Adrián; Garcia Gasulla, Dario; Béjar Alonso, Javier
The scarcity and imbalance of datasets for training deep learning models in a specific task is a common problem. This is especially true in the physiological domain where many applications use complex data collection processes and protocols, and it is difficult to gather a significant number of subjects. In this paper, we evaluate generative deep learning algorithms by training them to create data based on open physiological datasets and conduct a study on their potential for transfer learning. We measure the performance change of classifiers when the training data is augmented with the synthetic samples and also perform experiments in which we fine-tune classification models trained with the generated data adding increasing amounts of the real data to investigate the transfer learning capabilities of synthetic datasets. Finally, we advise and provide the best option for researchers interested in augmenting ECG datasets using these algorithms and the best fine-tuning strategies that would generalize correctly when tested on new data from the same domain but for a different classification task.
2024-03-20T06:46:04ZNúñez Rodríguez, José FernandoArjona Martínez, JamieTormos Llorente, AdriánGarcia Gasulla, DarioBéjar Alonso, JavierThe scarcity and imbalance of datasets for training deep learning models in a specific task is a common problem. This is especially true in the physiological domain where many applications use complex data collection processes and protocols, and it is difficult to gather a significant number of subjects. In this paper, we evaluate generative deep learning algorithms by training them to create data based on open physiological datasets and conduct a study on their potential for transfer learning. We measure the performance change of classifiers when the training data is augmented with the synthetic samples and also perform experiments in which we fine-tune classification models trained with the generated data adding increasing amounts of the real data to investigate the transfer learning capabilities of synthetic datasets. Finally, we advise and provide the best option for researchers interested in augmenting ECG datasets using these algorithms and the best fine-tuning strategies that would generalize correctly when tested on new data from the same domain but for a different classification task.Adaptive task-oriented chatbots using feature-based knowledge basesCampàs Gené, CarlaMotger de la Encarnación, JoaquimFranch Gutiérrez, JavierMarco Gómez, Jordihttp://hdl.handle.net/2117/4045092024-03-18T01:58:03Z2024-03-14T10:16:15ZAdaptive task-oriented chatbots using feature-based knowledge bases
Campàs Gené, Carla; Motger de la Encarnación, Joaquim; Franch Gutiérrez, Javier; Marco Gómez, Jordi
Task-oriented chatbots relying on a knowledge base for domain-specific content exploitation have been largely addressed in research and industry applications. Despite this, multiple challenges remain to be fully conquered, including adaptive knowledge mechanisms, personalization for user-specific demands, and composite intent resolution. To address these challenges, in this paper, we present a work-in-progress summary of a task-oriented, knowledge-based chatbot in the field of mobile software ecosystems. The chatbot is designed to assist users in the combined use of multiple features from different applications. The proposed knowledge base and the machine learning pipeline supporting the chatbot technical core are designed to: (i) effectively use user context, (ii) process runtime feedback, (iii) use user historical data, and (iv) automatically infer slot values and dependent actions. With this report, we expect to lay the groundwork for future development stages and user validation studies.
2024-03-14T10:16:15ZCampàs Gené, CarlaMotger de la Encarnación, JoaquimFranch Gutiérrez, JavierMarco Gómez, JordiTask-oriented chatbots relying on a knowledge base for domain-specific content exploitation have been largely addressed in research and industry applications. Despite this, multiple challenges remain to be fully conquered, including adaptive knowledge mechanisms, personalization for user-specific demands, and composite intent resolution. To address these challenges, in this paper, we present a work-in-progress summary of a task-oriented, knowledge-based chatbot in the field of mobile software ecosystems. The chatbot is designed to assist users in the combined use of multiple features from different applications. The proposed knowledge base and the machine learning pipeline supporting the chatbot technical core are designed to: (i) effectively use user context, (ii) process runtime feedback, (iii) use user historical data, and (iv) automatically infer slot values and dependent actions. With this report, we expect to lay the groundwork for future development stages and user validation studies.On the consistency of circuit lower bounds for non-deterministic timeAtserias, AlbertBuss, SamMüller, Moritzhttp://hdl.handle.net/2117/4035742024-03-03T15:43:15Z2024-03-01T09:03:15ZOn the consistency of circuit lower bounds for non-deterministic time
Atserias, Albert; Buss, Sam; Müller, Moritz
We prove the first unconditional consistency result for superpolynomial circuit lower bounds with a relatively strong theory of bounded arithmetic. Namely, we show that the theory V20 is consistent with the conjecture that NEXP ⊈ P/poly, i.e., some problem that is solvable in non-deterministic exponential time does not have polynomial size circuits. We suggest this is the best currently available evidence for the truth of the conjecture. Additionally, we establish a magnification result on the hardness of proving circuit lower bounds.
2024-03-01T09:03:15ZAtserias, AlbertBuss, SamMüller, MoritzWe prove the first unconditional consistency result for superpolynomial circuit lower bounds with a relatively strong theory of bounded arithmetic. Namely, we show that the theory V20 is consistent with the conjecture that NEXP ⊈ P/poly, i.e., some problem that is solvable in non-deterministic exponential time does not have polynomial size circuits. We suggest this is the best currently available evidence for the truth of the conjecture. Additionally, we establish a magnification result on the hardness of proving circuit lower bounds.On the pertinence of social practices for social roboticsClodic, AurélieVázquez Salceda, JavierDignum, FrankMascarenhas, SamuelDignum, VirginiaAugello, AgneseGentile, ManuelAlami, Rachidhttp://hdl.handle.net/2117/4026522024-02-26T00:26:41Z2024-02-22T12:12:01ZOn the pertinence of social practices for social robotics
Clodic, Aurélie; Vázquez Salceda, Javier; Dignum, Frank; Mascarenhas, Samuel; Dignum, Virginia; Augello, Agnese; Gentile, Manuel; Alami, Rachid
In the area of consumer robots that need to have rich social interactions with humans, one of the challenges is the complexity of computing the appropriate interactions in a cognitive, social and physical context. We propose a novel appproach for social robots based on the concept of Social Practices. By using social practices robots are able to be aware of their own social identities (given by the role in the social practice) and the identities of others and also be able to identify
the different social contexts and the appropriate social interactions that go along with those contexts and identities.
2024-02-22T12:12:01ZClodic, AurélieVázquez Salceda, JavierDignum, FrankMascarenhas, SamuelDignum, VirginiaAugello, AgneseGentile, ManuelAlami, RachidIn the area of consumer robots that need to have rich social interactions with humans, one of the challenges is the complexity of computing the appropriate interactions in a cognitive, social and physical context. We propose a novel appproach for social robots based on the concept of Social Practices. By using social practices robots are able to be aware of their own social identities (given by the role in the social practice) and the identities of others and also be able to identify
the different social contexts and the appropriate social interactions that go along with those contexts and identities.Design for values for social robot architecturesDignum, VirginiaDignum, FrankVázquez Salceda, JavierClodic, AurélieGentile, ManuelMascarenhas, SamuelAugello, Agnesehttp://hdl.handle.net/2117/4026482024-02-26T00:25:10Z2024-02-22T11:54:08ZDesign for values for social robot architectures
Dignum, Virginia; Dignum, Frank; Vázquez Salceda, Javier; Clodic, Aurélie; Gentile, Manuel; Mascarenhas, Samuel; Augello, Agnese
The integration of social robots in human societies requires that they are capable to take decisions that may affect the lives of people around them. In order to ensure that these robots will behave according to shared ethical principles, an important shift in the design and development of social robots is needed, one where the main goal is improving ethical transparency rather than technical performance, and placing human values at the core of robot designs. In this abstract, we discuss the concept of ethical decision making and how to achieve trust according to the principles of Autonomy, Responsibility and Transparency (ART).
2024-02-22T11:54:08ZDignum, VirginiaDignum, FrankVázquez Salceda, JavierClodic, AurélieGentile, ManuelMascarenhas, SamuelAugello, AgneseThe integration of social robots in human societies requires that they are capable to take decisions that may affect the lives of people around them. In order to ensure that these robots will behave according to shared ethical principles, an important shift in the design and development of social robots is needed, one where the main goal is improving ethical transparency rather than technical performance, and placing human values at the core of robot designs. In this abstract, we discuss the concept of ethical decision making and how to achieve trust according to the principles of Autonomy, Responsibility and Transparency (ART).