CRIS

Permanent URI for this communityhttps://scripta.up.edu.mx/handle/20.500.12552/1

Browse

Search Results

Now showing 1 - 6 of 6

Some of the metrics are blocked by your
consent settings
Pre-trained Models for Grammatical Error Correction in Healthcare-Specific Text
(Springer Nature Switzerland, 2025)
González Mora, José Guillermo
;
Ponce, Hiram
Grammatical Error Correction (GEC) is a common Natural Language Processing task and has been studied extensively in the field. There is an added complexity when trying to correct domain-specific text that is uncommon in general-purpose corpora. In this paper, we present a comparative study of three different approaches using pre-trained language models for Grammatical Error Correction in healthcare-specific text. We evaluated the performance of all proposals using the GLEU score metric, which allows a quantitative comparison of all methods. We utilized two different problem settings: first, a sequence-to-sequence pre-trained T5 model, followed by a fine-tuning process over a small set of examples. The second model is a pre-trained large language model, for which we test both zero-shot and few-shot in-context learning. The study shows how the fine-tuned T5 model is capable of exceeding the performance shown by the LLM tested. With this result, we show how smaller encoder-decoder models can solve domain-specific tasks with fewer parameters than a purely generative pre-trained LLM. ©The authors ©Springer.
Some of the metrics are blocked by your
consent settings
Contrastive Steering Vectors for Autoencoder Explainability
(MDPI, 2025)
González Mora, José Guillermo
;
Ponce, Hiram
;
Martinez-Villaseñor, Lourdes
Generative models, particularly autoencoders, often function as black boxes, making it challenging for non-expert users to effectively control the generation process and understand how inputs affect outputs. Existing methods for improving interpretability and control frequently require specific model training regimes or labeled data, limiting their applicability. This work introduces a novel approach to enhance the controllability and explainability of generative models, specifically tested on autoencoders with entangled latent spaces. We propose using a semi-supervised contrastive learning setup to learn steering vectors. These vectors, when added to an input’s latent representation, effectively manipulate specific attributes in the generated output without conditional training of the model or attribute classifiers, thus being applicable to pretrained models and avoiding compound classification errors. Furthermore, we leverage these learned steering vectors to interpret and explain the decoding process of a target attribute, allowing for efficient exploration of feature dimension interactions and the construction of an interpretable plot of the generative process, while lowering scalability limitations of perturbation-based Explainable AI (XAI) methods by reducing the search space. Our method provides an efficient pathway to controllable generation, offers an interpretable result of the model’s internal mechanisms, and relates the interpretations to human-understandable explanation questions. ©The authors ©MDPI AG ©Electronics.
Some of the metrics are blocked by your
consent settings
A Reinforcement Learning Method for Continuous Domains Using Artificial Hydrocarbon Networks
(2018)
Ponce, Hiram
;
González Mora, José Guillermo
;
Martinez-Villaseñor, Lourdes
Reinforcement learning in continuous states and actions has been limitedly studied in ocassions given difficulties in the determination of the transition function, lack of performance in continuous-to-discrete relaxation problems, among others. For instance, real-world problems, e.g. Fobotics, require these methods for learning complex tasks. Thus, in this paper, we propose a method for reinforcement learning with continuous states and actions using a model-based approach learned with artificial hydrocarbon networks (AHN). The proposed method considers modeling the dynamics of the continuous task with the supervised AHN method. Initial random rollouts and posterior data collection from policy evaluation improve the training of the AHN-based dynamics model. Preliminary results over the well-known mountain car task showed that artificial hydrocarbon networks can contribute to model-based approaches in continuous RL problems in both estimation efficiency (0.0012 in root mean squared-error) and sub-optimal policy convergence (reached in 357 steps), in just 5 trials over a parameter space θin R86. Data from experimental results are available at: http://sites.google.com/up.edu.mx/reinforcement-learning/ ©2018 IEEE.
Scopus© Citations 3 34 1
Some of the metrics are blocked by your
consent settings
An Implementation of a Monocular 360-Degree Vision System for Mobile Robot Navigation
(2018)
Acevedo Medina, Eduardo
;
Beltrán, Arturo
;
Castellanos Canales, Mauricio
;
Chaverra, Luis
;
González Mora, José Guillermo
One of the problems facing autonomous navigation is obstacle sensing and dynamic surroundings. Multi-sensor systems and omni directional vision systems have been implemented to increase the observability of robots. However, these approaches consider several drawbacks: the cost of processing multiple signals, synchronization of data collection, cost of materials and energy consumption, among others. Thus in this paper, we propose a new 360-degree vision system for mobile robot navigation using a static monocular camera. We demonstrate that using our system, it is possible to monitor the complete surroundings of the robot with a single sensor, i.e. the camera. Moreover, it can detect the position and orientation of an object from an egocentric point of view. We also present a low cost prototype of our proposal to validate it. © 2018 IEEE.
Scopus© Citations 3 25 1
Some of the metrics are blocked by your
consent settings
Development of Fast and Reliable Nature-Inspired Computing for Supervised Learning in High-Dimensional Data
(2019)
Ponce, Hiram
;
Souza, Paulo
;
González Mora, José Guillermo
Machine learning and data mining tasks in big data involve different nature of inputs that typically exhibit high dimensionality, e.g. more than 1,000 features, far from current acceptable scales computing in one machine. In many different domains, data have highly nonlinear representations that nature-inspired models can easily capture, outperforming simple models. But, the usage of these approaches in high-dimensional data are computationally costly. Recently, artificial hydrocarbon networks (AHN)—a supervised learning method inspired on organic chemical structures and mechanisms—have shown improvements in predictive power and interpretability in contrast with other well-known machine learning models, such as neural networks and random forests. However, AHN are very time-consuming that are not able to deal with big data until now. In this chapter, we present a fast and reliable nature-inspired training method for AHN, so they can handle high-dimensional data. This training method comprises a population-based meta-heuristic optimization with defined both individual encoding and objective function related to the AHN-model, and it is also implemented in parallel-computing. After benchmark performing of population-based optimization methods, grey wolf optimization (GWO) was selected. Our results demonstrate that the proposed hybrid GWO-based training method for AHN runs more than 1400x faster in high-dimensional data, without loss of predictability, yielding a fast and reliable nature-inspired machine learning model. We also present a use case in assisted living monitoring, i.e. human fall classification comprising 1,269 features from sensor signals and video recordings, with this proposed training algorithm to show its implementation and performance. We anticipate our new training algorithm to be useful in many applications like medical engineering, robotics, finance, aerospace, and others, in which big data is essential. © Springer Nature Switzerland AG 2020.
Scopus© Citations 3 15 2
Some of the metrics are blocked by your
consent settings
Stochastic parallel extreme artificial hydrocarbon networks: An implementation for fast and robust supervised machine learning in high-dimensional data
(2020)
De Campos Souza, Paulo V.
;
Junio Guimarães, Augusto
;
Ponce, Hiram
;
González Mora, José Guillermo
Artificial hydrocarbon networks (AHN) – a supervised learning method inspired on organic chemical structures and mechanisms – have shown improvements in predictive power and interpretability in comparison with other well-known machine learning models. However, AHN are very time-consuming that are not able to deal with large data until now. In this paper, we introduce the stochastic parallel extreme artificial hydrocarbon networks (SPE-AHN), an algorithm for fast and robust training of supervised AHN models in high-dimensional data. This training method comprises a population-based meta-heuristic optimization with defined individual encoding and objective function related to the AHN-model, an implementation in parallel-computing, and a stochastic learning approach for consuming large data. We conducted three experiments with synthetic and real data sets to validate the training execution time and performance of the proposed algorithm. Experimental results demonstrated that the proposed SPE-AHN outperforms the original-AHN method, increasing the speed of training more than 10,000x times in the worst case scenario. Additionally, we present two case studies in real data sets for solar-panel deployment prediction (regression problem), and human falls and daily activities classification in healthcare monitoring systems (classification problem). These case studies showed that SPE-AHN improves the state-of-the-art machine learning models in both engineering problems. We anticipate our new training algorithm to be useful in many applications of AHN like robotics, finance, medical engineering, aerospace, and others, in which large amounts of data (e.g. big data) is essential. © 2019 Elsevier Ltd.
Scopus© Citations 21 10 5

CRIS

Browse

Filters

Settings

Sort By

Results per page

Search Results