Interpretable Solutions for Breast Cancer Diagnosis with Grammatical Evolution and Data Augmentation
Authors:
Yumnah Hasan,
Allan de Lima,
Fatemeh Amerehi,
Darian Reyes Fernandez de Bulnes,
Patrick Healy,
Conor Ryan
Abstract:
Medical imaging diagnosis increasingly relies on Machine Learning (ML) models. This is a task that is often hampered by severely imbalanced datasets, where positive cases can be quite rare. Their use is further compromised by their limited interpretability, which is becoming increasingly important. While post-hoc interpretability techniques such as SHAP and LIME have been used with some success on…
▽ More
Medical imaging diagnosis increasingly relies on Machine Learning (ML) models. This is a task that is often hampered by severely imbalanced datasets, where positive cases can be quite rare. Their use is further compromised by their limited interpretability, which is becoming increasingly important. While post-hoc interpretability techniques such as SHAP and LIME have been used with some success on so-called black box models, the use of inherently understandable models makes such endeavors more fruitful. This paper addresses these issues by demonstrating how a relatively new synthetic data generation technique, STEM, can be used to produce data to train models produced by Grammatical Evolution (GE) that are inherently understandable. STEM is a recently introduced combination of the Synthetic Minority Oversampling Technique (SMOTE), Edited Nearest Neighbour (ENN), and Mixup; it has previously been successfully used to tackle both between class and within class imbalance issues. We test our technique on the Digital Database for Screening Mammography (DDSM) and the Wisconsin Breast Cancer (WBC) datasets and compare Area Under the Curve (AUC) results with an ensemble of the top three performing classifiers from a set of eight standard ML classifiers with varying degrees of interpretability. We demonstrate that the GE-derived models present the best AUC while still maintaining interpretable solutions.
△ Less
Submitted 25 January, 2024;
originally announced January 2024.
Monte Carlo study of multicomponent monolayer adsorption on square lattices
Authors:
Guillermo D. García,
Fabricio O. Sánchez-Varretti,
Fernando Bulnes,
Antonio J. Ramirez-Pastor
Abstract:
The monolayer adsorption process of interacting binary mixtures of species $A$ and $B$ on square lattices is studied through grand canonical Monte Carlo simulation in the framework of the lattice-gas model. Four different energies have been considered in the adsorption process: 1) $ε_0$, interaction energy between a particle (type $A$ or $B$) and a lattice site; 2) $w_{AA}$, interaction energy bet…
▽ More
The monolayer adsorption process of interacting binary mixtures of species $A$ and $B$ on square lattices is studied through grand canonical Monte Carlo simulation in the framework of the lattice-gas model. Four different energies have been considered in the adsorption process: 1) $ε_0$, interaction energy between a particle (type $A$ or $B$) and a lattice site; 2) $w_{AA}$, interaction energy between two nearest-neighbor $A$ particles; 3) $w_{BB}$, interaction energy between two nearest-neighbor $B$ particles; 4) $w_{AB}=w_{BA}$, interaction energy between two nearest-neighbors being one of type $A$ and the other of type $B$. The adsorption process has been monitored through total and partial isotherms and differential heats of adsorption corresponding to both species of the mixture. Our main interest is in the repulsive lateral interactions, where a variety of structural orderings arise in the adlayer, depending on the interaction parameters ($w_{AA}$, $w_{BB}$ and $w_{AB}$). At the end of this work, we determine the phase diagram characterizing the phase transitions occurring in the system. A nontrivial interdependence between the partial surface coverage of both species is observed.
△ Less
Submitted 30 May, 2011;
originally announced May 2011.
