# datascience

Sebastian Kwiatkowski

01/10/2018, 6:53 AM
Hi, I'm the author of Komputation. I'm late to the party, but I wanted to respond to @czyzby's criticism (from November 28). The layer definitions have been simplified. The example that was quoted now reads:
convolution(numberFilters, filterWidth, filterHeight, initialization, optimization)
Komputation itself won't provide too many default layer settings. That seems like a half-baked solution. Appropriate choices depend on the context. I do have plans to explore neural architecture search features, though.