From 24c69c75e2320004edc8dc6c8dd0e2a2a8119390 Mon Sep 17 00:00:00 2001 From: ravenscroftj Date: Sun, 23 Jan 2022 17:55:03 +0000 Subject: [PATCH] Update 'brainsteam/data/mentions.json' --- brainsteam/data/mentions.json | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) diff --git a/brainsteam/data/mentions.json b/brainsteam/data/mentions.json index 84fa2cf..1ff168e 100644 --- a/brainsteam/data/mentions.json +++ b/brainsteam/data/mentions.json @@ -1789,6 +1789,25 @@ "content": "

@jamesravey chaotic systems! when i was doing speech rec training we would regularly get mandates from the boss or the customer to \"improve the frontier graph\" by x points or whatever and they'd grill us on what we thought would give that improvement. about I dunno like 50% of the time the result of executing that plan was a degradation of model performance. (this is why I left that gig too, the catch-22 of improving already-optimized chaotic systems was too much)

", "published": "2022-01-23T16:39:42" } + }, + { + "id": 1336696, + "source": "https://brid.gy/comment/mastodon/@jamesravey@fosstodon.org/107672683821046429/107672941340348101", + "target": "https://brainsteam.co.uk/2022/01/22/deep-learning-reproducibility-here-be-dragons/", + "activity": { + "type": "reply" + }, + "verified_date": "2022-01-23T12:55:02.573304", + "data": { + "author": { + "type": "card", + "name": "James Ravenscroft", + "photo": "https://webmention.io/avatar/cdn.fosstodon.org/1db69bf6408d549f11885537e17a22d22c58563968b1d6462a3f8e6d47619670.png", + "url": "https://fosstodon.org/@jamesravey" + }, + "content": "

@capn_b ah that old chestnut! I guess it's almost impossible to avoid overfitting when you're already performing well on general use cases

", + "published": "2022-01-23T17:24:59" + } } ] } \ No newline at end of file