With the shrinking of the IC technology node, optical proximity effects (OPC) and etch proximity effects (EPC) are the two major tasks in advanced photolithography patterning. Machine learning has emerged in OPC/EPC problems because conventional optical-solver-based OPC is time-consuming, and there is no physical model existing for EPC. In this work, we use dimensionality reduction (DR) algorithms to reduce the computation time of complex OPC/EPC problems while the prediction accuracy is maintained. Also, we implement a pure machine learning approach where the input masks are directly mapped to the output etched patterns. While one photolithographic mask can generate many experimental patterns at once, our pure ML-based approach can shorten the trial-and-error period in the photolithographic correction. Additionally, we demonstrate the automation in SEM images preprocessing using feature detection, and this facilitates intelligent manufacturing in semiconductor processing. The input vector dimensions are effectively reduced by two orders of magnitude while the observed mean squared error is not affected significantly. The computation runtime is reduced from 4804 s of the baseline calculation to 10 s-200 s The MSE values changed from the baseline 0.037 to 0.037 for singular value decomposition (SVD), to 0.039 for independent component analysis (ICA), and to 0.035 for factor analysis (FA).

Several computational approaches for predicting subcellular localization have been developed and proposed. These approaches provide diverse performance because of their different combinations of protein features, training datasets, training strategies, and computational machine learning algorithms. In some cases, these tools may yield inconsistent and conflicting prediction results. It is important to consider such conflicting or contradictory predictions from multiple prediction programs during protein annotation, especially in the case of a multiclass classification problem such as subcellular localization. Hence, to address this issue, this work proposes the use of the particle swarm optimization (PSO) algorithm to combine the prediction outputs from multiple different subcellular localization predictors with the aim of integrating diverse prediction models to enhance the final predictions. Herein, we present PSO-LocBact, a consensus classifier based on PSO that can be used to combine the strengths of several preexisting protein localization predictors specially designed for bacteria. Our experimental results indicate that the proposed method can resolve inconsistency problems in subcellular localization prediction for both Gram-negative and Gram-positive bacterial proteins. The average accuracy achieved on each test dataset is over 98%, higher than that achieved with any individual predictor.

No abstract

No abstract