Scientists rise up against statistical significance | Nature
For several generations, researchers have been warned that a statistically non-significant result does not ‘prove’ the null hypothesis (the hypothesis that there is no difference between groups or no effect of a treatment on some measured outcome). Nor do statistically significant results ‘prove’ some other hypothesis. Such misconceptions have famously warped the literature with overstated claims and, less famously, led to claims of conflicts between studies where none exists.
Cell composition analysis of bulk genomics using single-cell data | Nature Methods
Single-cell RNA sequencing (scRNA-seq) is a rich resource of cellular heterogeneity, opening new avenues in the study of complex tissues. Researchers introduce Cell Population Mapping (CPM), a deconvolution algorithm in which reference scRNA-seq profiles are leveraged to infer the composition of cell types and states from bulk transcriptome data (‘scBio’ CRAN R-package).
Measuring the Limits of Data Parallel Training for Neural Networks | Google AI Blog
In “Measuring the Effects of Data Parallelism in Neural Network Training”, researchers investigate the relationship between batch size and training time by running experiments on six different types of neural networks across seven different datasets using three different optimization algorithms (“optimizers”). In total, they trained over 100K individual models across ~450 workloads, and observed a seemingly universal relationship between batch size and training time across all workloads they tested.