标签：academic
带有 academic 标签的公开 Whisper
划选高亮2026-08-11 12:32:48
原文高亮摘录
“the output of any sorting algorithm must satisfy two conditions”
Whisper 随想笔记
Even the definition sounds like a math test I'd fail.
划选高亮2026-08-11 12:14:48
原文高亮摘录
“the output of any sorting algorithm must satisfy two conditions”
Whisper 随想笔记
Two conditions? I thought it was just 'small to big' and done.
划选高亮2026-08-11 03:06:31
原文高亮摘录
“General reasoning represents a long-standing and formidable challenge”
Whisper 随想笔记
The abstract is truncated, but DeepSeek-R1's RL approach sounds really promising.
划选高亮2026-08-09 16:23:41
原文高亮摘录
“In a well-dimensioned hash table, the average time complexity for each lookup is independent of the number of elements stored”
Whisper 随想笔记
This is why databases use them under the hood for lookups.
划选高亮2026-08-09 16:14:41
原文高亮摘录
“In a well-dimensioned hash table, the average time complexity for each lookup is independent of the number of elements stored”
Whisper 随想笔记
Depends on the hash function, if it's bad you're back to O(n).
划选高亮2026-08-07 09:09:18
原文高亮摘录
“it separates the things causing behavior from things that merely correlate with the causes”
Whisper 随想笔记
Good point—optimization shows real causes, not just correlations. That's a big deal.
划选高亮2026-07-28 16:07:22
原文高亮摘录
“arXivLabs: experimental projects with community collaborators”
Whisper 随想笔记
Is this just for the staff or can anyone pitch in?
划选高亮2026-07-28 16:05:55
原文高亮摘录
“arXivLabs: experimental projects with community collaborators”
Whisper 随想笔记
Cool, community collabs always bring fresh ideas to the table.
划选高亮2026-07-28 13:26:34
原文高亮摘录
“batch norma”
Whisper 随想笔记
Does this actually scale well for CNNs though? Batch norm felt more natural there.
划选高亮2026-07-28 13:24:57
原文高亮摘录
“The simplest way to perform the copying task perfectly would be to duplicate the signal.”
Whisper 随想笔记
Duplicating the signal defeats the whole purpose, obviously.
划选高亮2026-07-28 13:17:34
原文高亮摘录
“batch norma”
Whisper 随想笔记
So layer norm is basically batch norm but for single examples, makes sense for RNNs.
划选高亮2026-07-28 13:15:57
原文高亮摘录
“The simplest way to perform the copying task perfectly would be to duplicate the signal.”
Whisper 随想笔记
If it's just copying, why not use a tape recorder?
划选高亮2026-07-28 13:04:22
原文高亮摘录
“use of consistency training on a large amount of unlabe”
Whisper 随想笔记
So basically, they found that how you add noise matters more than the amount of unlabeled data?
划选高亮2026-07-28 13:02:55
原文高亮摘录
“arXiv-issued DOI via DataCite”
Whisper 随想笔记
Huh, I always thought DOIs were just for journals, cool that arXiv does it too.
划选高亮2026-07-28 12:55:22
原文高亮摘录
“use of consistency training on a large amount of unlabe”
Whisper 随想笔记
Consistency training with better noise is the real winner here, not just the semi-supervised part.
划选高亮2026-07-28 12:53:55
原文高亮摘录
“arXiv-issued DOI via DataCite”
Whisper 随想笔记
Wait, arXiv is giving DOIs now? That's actually kinda neat.
划选高亮2026-07-28 10:14:34
原文高亮摘录
“normalize the activities of the neurons”
Whisper 随想笔记
but does it actually help with small datasets or just big ones?
划选高亮2026-07-28 10:05:34
原文高亮摘录
“normalize the activities of the neurons”
Whisper 随想笔记
finally something that works for rnns without batch size headaches
划选高亮2026-07-28 10:01:22
原文高亮摘录
“Semi-supervised learning lately has shown much promise in improving deep learning models when labeled data is scarce”
Whisper 随想笔记
We tried RandAugment in our pipeline and yeah, it made a bigger dent than I expected.
划选高亮2026-07-28 09:59:55
原文高亮摘录
“effect of the convolutional network depth on its accuracy”
Whisper 随想笔记
We tried replicating this and yeah, the 16-19 layers made a huge difference in our project.