search

tyler-tomita / RandomerForest

Discriminant Projection Forest results, datasets, etc.
44 stars 21 forks source link

let's add feature selection to this manuscript, #105

Open jovo opened 7 years ago

jovo commented 7 years ago

either pre or post submission, but lots of people ask about it immediately after asking about performance.

i think we could simply count the number of times a feature is used, and/or the number of times a subspace is used.

tyler-tomita commented 7 years ago

i will add that to the outline

tyler-tomita commented 7 years ago

feature importance is easy because the set has cardinality p. it's not clear to me what a feasible way to plot subspace importance is because the set of all subspaces is vastly larger than the set of all features

jovo commented 7 years ago

we can just list the top 10 (say).

On Tue, Jan 10, 2017 at 1:43 PM, Tyler Tomita notifications@github.com wrote:

feature importance is easy because the set has cardinality p. it's not clear to me what a feasible way to plot subspace importance is because the set of all subspaces is vastly larger than the set of all features

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ttomita/RandomerForest/issues/105#issuecomment-271660625, or mute the thread https://github.com/notifications/unsubscribe-auth/AACjcmtiTrkQ_ry6tHmdiKZFwKiMmaXAks5rQ9FcgaJpZM4Lfbmk .

-- the glass is all full: half water, half air. neurodata.io, jovo calendar https://calendar.google.com/calendar/embed?src=joshuav%40gmail.com&ctz=America/New_York starting in 2017, i will be responding to emails approximately weekly, so that i can focus more on the people around me. thank you for your understanding.