Conversation

Notices

1. Embed this notice
   phiofx (phiofx@hachyderm.io)'s status on Sunday, 21-Jan-2024 19:34:00 JST phiofx

   in reply to

   • MissMythreyi

   @missmythreyi you cant create something out of nothing. But making optimal use of scarce information may be facilitated by validating ideas using datasets that are not limited by sample size.

   To put it otherwise: if your algorithm, tool or whatever doesnt perform with plenty of synthetic data, it wont perform with scarce real data.

   • Embed this notice
     MissMythreyi (missmythreyi@mastodon.world)'s status on Sunday, 21-Jan-2024 19:34:10 JST MissMythreyi

     Hello, what is 'synthetic data'? Specifically, how can synthetic data be sufficient for creating models to run on real-world data? Would it not replicate the narrowness of the small real data they had to begin with?

     Do I know anyone on here who explored what kind of niches synthetic data helps & where its use is being over-shot beyond the effective boundary?

     Tagging to hope to reach to someone who worked on this #machinelearning