In yeast, cells commit to division at Start in G1 phase. The critical cell size for Start is regulated in part by the size-control genes CLN3 and WHI5. Cln3 is a G1 cyclin that activates Start by activating the Cdc28 cyclin dependent kinase. Null cln3 mutants have large cells, because they lack an activator, while hyperactive mutants have small cells. Whi5 inhibits Start. Null whi5 mutants have small cells, while hyperactive mutants have large cells. Previous studies suggested that CLN3 and WHI5 are in a linear, dependent pathway, in which Cln3 acts inhibits Whi5, possibly by phosphorylation. We compared isogenic WT, cln3, whi5, and cln3 whi5 mutants, and find that whi5 is not epistatic, or not fully epistatic, to CLN3. That is, CLN3 can work independently of WHI5. The cln3 and whi5 deletions have off-setting phenotypes, such that the double mutant is nearly wild-type in cell size, and has only the mildest phenotype, with well-controlled cell size, despite entirely lacking two major cell size regulators. Growth rates were also measured. Some mutants were larger and some mutants smaller than wild-type, but the wild-type had the fastest doubling time, consistent with the idea that wild-type size is optimal for growth rate.
The Hidden Power of Normalization: Exponential Capacity Control in Deep Neural Networks
arXiv:2511.00958v1 Announce Type: cross Abstract: Normalization methods are fundamental components of modern deep neural networks (DNNs). Empirically, they are known to stabilize optimization dynamics and

