We carry out a critical assessment of machine learning and deep learning models for the classification of skin tumors. Machine learning (ML) algorithms tested in this work include logistic regression, linear discriminant analysis, k-nearest neighbors classifier, decision tree classifier and Gaussian naive Bayes, while deep learning (DL) models employed are either based on a custom Convolutional Neural Network model, or leverage transfer learning via the use of pre-trained models (VGG16, Xception and ResNet50). We find that DL models, with accuracies up to 0.88, all outperform ML models. ML models exhibit accuracies below 0.72, which can be increased to up to 0.75 with ensemble learning. To further assess the performance of DL models, we test them on a larger and more imbalanced dataset. Metrics, such as the F-score and accuracy, indicate that, after fine-tuning, pre-trained models perform extremely well for skin tumor classification. This is most notably the case for VGG16, which exhibits an F-score of 0.88 and an accuracy of 0.88 on the smaller database, and metrics of 0.70 and 0.88, respectively, on the larger database.
We investigate the dependence of the displacements of a molecular motor embedded inside a glassy material on its folding characteristic time τ f . We observe two different time regimes. For slow foldings (regime I) the diffusion evolves very slowly with τ f , while for rapid foldings (regime II) the diffusion increases strongly with τ f ( D ≈ τ −2 f ) suggesting two different physical mechanisms. We find that in regime I the motor's displacement during the folding process is counteracted by a reverse displacement during the unfolding, while in regime II this counteraction is much weaker. We notice that regime I behavior is reminiscent of the scallop theorem that holds for larger motors in a continuous medium. We find that the difference in the efficiency of the motor's motion explains most of the observed difference between the two regimes. For fast foldings the motor trajectories differ significantly from the opposite trajectories induced by the following unfolding process, resulting in a more efficient global motion than for slow foldings. This result agrees with the fluctuation theorems expectation for time reversal mechanisms. In agreement with the fluctuation theorems we find that the motors are unexpectedly more efficient when they are generating more entropy, a result that can be used to increase dramatically the motor's motion.
Using molecular simulation, we shed light on the crystal nucleation process in systems of Cu, Ni, and their nanoalloy. For each system, we simulate the formation of the crystal nucleus along the entire nucleation pathway and determine the free energy barrier overcome by the system to form a critical nucleus. Comparing the results obtained for the pure metals to those for the nanoalloy, we analyze the impact of alloying on the free energy of nucleation, as well as on the size and structure of the crystal nucleus. Specifically, we relate the greater free energy of nucleation, and bigger critical nuclei, obtained for the nanoalloy, to the difference in size and cohesive energy between the two metals. Furthermore, we characterize the dependence of the local composition of the incipient crystal cluster on its size, which is of key significance for the applications of bimetallic nanoparticles in catalysis.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.