On Mon, May 20, 2024 at 08:06 Undescribed Horrific Abuse, One Victim & Survivor of Many <[email protected]> wrote:
> > > On Mon, May 20, 2024 at 07:16 Undescribed Horrific Abuse, One Victim & > Survivor of Many <[email protected]> wrote: > >> >> >> On Mon, May 20, 2024 at 07:07 Undescribed Horrific Abuse, One Victim & >> Survivor of Many <[email protected]> wrote: >> >>> # we improved the stability of low-data machine learning model training >>> >>> ## why is this useful? >>> >>> low-data model training would let individual users make AI models >>> without access to big data, such as those models used at scale to addict >>> users to products, services, or behaviors. >>> >>> ability to recreate these models with the user in control would give >>> them many more options for personal freedom. >>> >>> additionally low-data training makes more powerful AI. >>> >> >> <snip> >> >> ## in short >>> >>> to make a big model on a small dataset, train a training model that >>> improves generalization to more data. >>> >>> this is more powerful but more difficult if it also improves >>> generalization of itself. >>> >> >> additionally: this may have been tried in research papers somewhere >> already, if attempting it is unreasonable. i might start with keywords >> around hyper-parameter metamodel for generalization or such >> > > uhhhh maybe near ideas of training the training model in an RL way such > that it hand-selects every batch (optionally every learning rate) > near online training, data curation sounds like it exists >
