Additional file 7 of Systematic interrogation of mutation groupings reveals divergent downstream expression programs within key cancer genes

Additional file 7: Figure S6. Increasing computational complexity does not change or improve upon classification performance. We observed similar subgrouping classification performance in METABRIC(LumA) when we repeated our prediction tasks with (a) a support vector machine classifier and (b) a random forest classifier in place of the logistic ridge regression classifier that was originally used. (c) Using these more computationally complex classifiers did not result in improved classification performance across all non-random classification tasks, nor did tuning using a greater number of potential hyper-parameter values.补充文件7：图S6。提升计算复杂度并未改变或优化分类性能。我们分别使用(a)支持向量机分类器（support vector machine classifier）与(b)随机森林分类器（random forest classifier）替换原实验所采用的逻辑岭回归分类器，重复预测任务后，在METABRIC队列的LumA亚型中观测到了相似的亚组分类性能。(c) 这类计算复杂度更高的分类器，并未在所有非随机分类任务中实现分类性能的提升；即便使用更多潜在超参数值进行调参，同样未能改善分类性能。