RT info:eu-repo/semantics/article
T1 Adjacent inputs with different labels and hardness in supervised learning
A1 Grillo, Sebastián Alberto
A1 Mello Román, Julio César
A1 Mello Román, Jorge Daniel
A1 Vázquez Noguera, José Luis
A1 García Torres, Miguel
A1 Divina, Federico
A1 Gardel Sotomayor, Pedro Esteban
A2 Universidad Americana/ INCADE S.A.E
K1 Complexity theory
K1 Data models
K1 Machine learning algorithms
K1 Measurement uncertainty
K1 Neural networks
K1 Particle measurements
K1 Supervised learning
AB An important aspect of the design of effective machine learning algorithms is the complexity analysis of classification problems. In this paper, we propose a study aimed at determining the relation between the number of adjacent inputs with different labels and the required number of examples for the task of inducing a classification model. To this aim, we first quantified the adjacent inputs with different labels as a property, using a measure denoted as Neighbour Input Variation (NIV). We analyzed the relation that NIV has to random data and overfitting. We then demonstrated that a threshold of NIV may determine if a classification model can generalize to unseen data. We also presented a case study aimed at analyzing threshold neural networks and the required first hidden layer size in function of NIV. Finally, we performed experiments with five popular algorithms analyzing the relation between NIV and the classification error on problems with few dimensions. We conclude that functions whose similar inputs have different outputs with high probability, considerably reduce the generalization capacity of classification algorithms.
PB Institute of Electrical and Electronics Engineers
YR 2021
FD 2021-11-25
LK http://hdl.handle.net/20.500.14066/4519
UL http://hdl.handle.net/20.500.14066/4519
LA eng
NO Grillo, S. A., Mello Román, J. C., Mello Román, J. D., Vázquez Noguera, J. L., García Torres, M., & Divina, F. (2021). Adjacent inputs with different labels and hardness in supervised learning. IEEE Access, 9, 162487-162498.https://doi.org/10.1109/ACCESS.2021.3131150
NO Corresponding author: Sebastián A. Grillo (sebastian.grillo@ua.edu.py)
NO Consejo Nacional de Ciencia y Tecnología
DS MINDS@UW
RD 05-feb-2025