mstdn.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A general-purpose Mastodon server with a 500 character limit. All languages are welcome.

Administered by:

Server stats:

11K
active users

#DataAugmentation

0 posts0 participants0 posts today
Kathy Reid<p>"Good luck, Viv, I know that guy's a total douche."</p><p>"Thanks Tay. Have you got the multi-wake word model training?"</p><p>"Yeah. Are you're sure he'll pick that <a href="https://aus.social/tags/WakeWord" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>WakeWord</span></a> though?"</p><p>"I'm pretty sure. He won't outright use the phrase "dole bludger" but it's pretty close."</p><p>---</p><p>A sardonic smile crept over the Prime Minister's face, the hot summer sun reflecting off his near-bald temples. </p><p>"So, by choosing a Wake Word that has difficult to pronounce sounds in it, it means that it won't work well for people who speak with an accent?"</p><p>"Yes, Prime Minister, precisely".</p><p>"Like the 'th' sound in them, there that?"</p><p>"Yes, or words that start or end with hard consonants are also difficult for some accents". </p><p>"Do we know which Wake Word would be the hardest for immigrants? Indigenous people?". </p><p>Pained, the <a href="https://aus.social/tags/linguist" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>linguist</span></a> had feared this question. She knew exactly what he was trying to do. </p><p>The long call centre queues hadn't done the trick - people had added screeners to their phones so that after being on hold for 7 hours, the phone would alert them to a picked up call. </p><p>The Assistant was supposed to be the replacement for the call centre. Just load the app on your phone, and ask it a question! So simple! No queuing! The government actually wanted to help people!</p><p>You just needed to use the Wake Word to "wake up" the assistant first. </p><p>"Well ---", she hesitated. </p><p>"I don't have all day Doctor!" </p><p>"Our research shows that a Wake Word like `This Starts With Me` has lots of hard to pronounce phonemes - sounds". </p><p>"Excellent. And I like the overtone of personal responsibility."</p><p>Of course the fucker did. </p><p>"Very well Prime Minister, we will implement that Wake Word". </p><p>He trotted off, probably to kill some babies or kittens, she thought.</p><p>---</p><p>Tay was configuring the <a href="https://aus.social/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a> for the <a href="https://aus.social/tags/ML" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ML</span></a> training run for the <a href="https://aus.social/tags/WakeWord" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>WakeWord</span></a> model. </p><p>They had just finished downloading every instance of the words "this", "starts" "with" "me" in every accent of English, from Common Voice. </p><p>By augmenting the Wake Word model with accent data, they could make it recognise more accents, more accurately. </p><p>Resistance came in many forms. </p><p>---<br><a href="https://arxiv.org/abs/2104.01454" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="">arxiv.org/abs/2104.01454</span><span class="invisible"></span></a><br><a href="https://dl.acm.org/doi/10.1145/3617694.3623258" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">dl.acm.org/doi/10.1145/3617694</span><span class="invisible">.3623258</span></a><br>---</p><p><a href="https://aus.social/tags/Tootfic" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Tootfic</span></a> <a href="https://aus.social/tags/Microfiction" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>Microfiction</span></a></p>
Python Torino<p>"Beer, Data &amp; Robots" ⚛️ è stata davvero una serata esplosiva.. grazie a tutti !! 💥 🙏🏻</p><p>Grazie a Simona Mazzarino e a Andrea Marchese per averci illustrato fino a che punto una IA può mal interpretare i dati e come poterli esplorare con un visore di realtà aumentata 📊</p><p>Ecco il video della serata: <a href="https://video.linux.it/w/3cnKaZqmSLtpEgkYc8jAFq" rel="nofollow noopener" target="_blank"><span class="invisible">https://</span><span class="ellipsis">video.linux.it/w/3cnKaZqmSLtpE</span><span class="invisible">gkYc8jAFq</span></a></p><p><a href="https://social.python.it/tags/databeerstorino" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>databeerstorino</span></a> <a href="https://social.python.it/tags/ai" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ai</span></a> <a href="https://social.python.it/tags/AIbias" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AIbias</span></a> <a href="https://social.python.it/tags/AIfairness" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AIfairness</span></a> <a href="https://social.python.it/tags/syntheticdata" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>syntheticdata</span></a> <a href="https://social.python.it/tags/XGBoost" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>XGBoost</span></a> <a href="https://social.python.it/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a> <a href="https://social.python.it/tags/vr" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>vr</span></a> <a href="https://social.python.it/tags/dataframe" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>dataframe</span></a> <a href="https://social.python.it/tags/open3d" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>open3d</span></a> <a href="https://social.python.it/tags/unity" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>unity</span></a> <a href="https://social.python.it/tags/pythontorino" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>pythontorino</span></a> <a href="https://social.python.it/tags/datascience" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>datascience</span></a> <a href="https://social.python.it/tags/python" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>python</span></a></p>
whitone<p>"Know your Bias: Tackling Data Bias through Synthetic Data" by Simona Mazzarino <a href="https://mastodon.social/tags/ClearboxAI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ClearboxAI</span></a> <a href="https://mastodon.social/tags/databeerstorino" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>databeerstorino</span></a> <span class="h-card" translate="no"><a href="https://social.python.it/@torino" class="u-url mention" rel="nofollow noopener" target="_blank">@<span>torino</span></a></span> — Beer, Data &amp; Robots</p><p><a href="https://mastodon.social/tags/ai" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ai</span></a> <a href="https://mastodon.social/tags/AIbias" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AIbias</span></a> <a href="https://mastodon.social/tags/AIfairness" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AIfairness</span></a> <a href="https://mastodon.social/tags/syntheticdata" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>syntheticdata</span></a> <a href="https://mastodon.social/tags/XGBoost" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>XGBoost</span></a> <a href="https://mastodon.social/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a></p>
InfoQ<p>Listen to the <a href="https://techhub.social/tags/InfoQ" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>InfoQ</span></a> <a href="https://techhub.social/tags/podcast" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>podcast</span></a> featuring Sam Partee, where he shares insights on Redis' vector database offering, different approaches to embeddings, and how to enhance <a href="https://techhub.social/tags/LLMs" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>LLMs</span></a> by adding a search component for retrieval augmented generation: <a href="https://bit.ly/3ukrEjw" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="">bit.ly/3ukrEjw</span><span class="invisible"></span></a> </p><p>Plus, a peek into the world of hybrid search in Redis! </p><p><a href="https://techhub.social/tags/AI" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>AI</span></a> <a href="https://techhub.social/tags/ML" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ML</span></a> <a href="https://techhub.social/tags/DataBase" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataBase</span></a> <a href="https://techhub.social/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a></p>
Amy Tabb 🇺🇦<p>Soft Augmentation for Image Classification<br> Authors: Yang Liu, Shen Yan, Laura Leal-Taixé, James Hays, Deva Ramanan </p><p>abs: <a href="http://arxiv.org/abs/2211.04625" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">http://</span><span class="">arxiv.org/abs/2211.04625</span><span class="invisible"></span></a> <br>code: <a href="https://github.com/youngleox/soft_augmentation" rel="nofollow noopener" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">github.com/youngleox/soft_augm</span><span class="invisible">entation</span></a> </p><p><a href="https://hachyderm.io/tags/arXiv" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>arXiv</span></a> <a href="https://hachyderm.io/tags/ComputerVision" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>ComputerVision</span></a> <a href="https://hachyderm.io/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a></p>
Dan Stowell<p>People applying <a href="https://mastodon.social/tags/deeplearning" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>deeplearning</span></a> to <a href="https://mastodon.social/tags/bioacoustics" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>bioacoustics</span></a> (animal sounds) often ask me about <a href="https://mastodon.social/tags/dataaugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>dataaugmentation</span></a> strategies. Here's my answer, on Stack Exchange <a href="https://bioacoustics.stackexchange.com/questions/98/data-augmentation-strategies-for-bioacoustics-machine-learning" rel="nofollow noopener" target="_blank"><span class="invisible">https://</span><span class="ellipsis">bioacoustics.stackexchange.com</span><span class="invisible">/questions/98/data-augmentation-strategies-for-bioacoustics-machine-learning</span></a></p>
Marcin Paprzycki<p>Enhancing the quality of a small and unbalanced dataset by use of preprocessing and augmentation methods in: “Treating Dataset Imbalance in Fetal Echocardiography Classification” by Guilherme Gusmão, Alberto Raposo, Renato de Oliveira, Carlos Barbosa. Communication Papers of the 17th Conference on Computer Science and Intelligence Systems; ACSIS, Vol. 32, pages 3–9 (2022).</p><p><a href="https://masto.ai/tags/fetalechocardiography" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>fetalechocardiography</span></a> <a href="https://masto.ai/tags/dataaugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>dataaugmentation</span></a> <a href="https://masto.ai/tags/imageprocessing" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>imageprocessing</span></a><br>Open Access: <a href="https://lnkd.in/dwuYikMf" rel="nofollow noopener" target="_blank"><span class="invisible">https://</span><span class="">lnkd.in/dwuYikMf</span><span class="invisible"></span></a></p>
Hiram Ring<p>tldr; <a href="https://mstdn.social/tags/data" class="mention hashtag" rel="tag">#<span>data</span></a> <a href="https://mstdn.social/tags/augmentation" class="mention hashtag" rel="tag">#<span>augmentation</span></a> in <a href="https://mstdn.social/tags/NLProc" class="mention hashtag" rel="tag">#<span>NLProc</span></a> degrades <a href="https://mstdn.social/tags/textClassification" class="mention hashtag" rel="tag">#<span>textClassification</span></a> performance in most cases. <a href="https://aclanthology.org/2022.insights-1.12.pdf" target="_blank" rel="nofollow noopener" translate="no"><span class="invisible">https://</span><span class="ellipsis">aclanthology.org/2022.insights</span><span class="invisible">-1.12.pdf</span></a></p><p>Well that was fun. Just spent last night experimenting with <a href="https://mstdn.social/tags/dataAugmentation" class="mention hashtag" rel="tag">#<span>dataAugmentation</span></a> (using <a href="https://mstdn.social/tags/flan" class="mention hashtag" rel="tag">#<span>flan</span></a>-T5 for paraphrasing) and in wondering why it seems to degrade <a href="https://mstdn.social/tags/textClassification" class="mention hashtag" rel="tag">#<span>textClassification</span></a> performance I came across this great paper essentially saying the same thing. I guess I’ll revisit this in a year or so when there are better language models.<br /><span class="h-card" translate="no"><a href="https://a.gup.pe/u/linguistics" class="u-url mention">@<span>linguistics</span></a></span> <span class="h-card" translate="no"><a href="https://sigmoid.social/about/more?instance_actor=true" class="u-url mention">@<span>sigmoid.social</span></a></span></p>
Solal Nathan<p>What do you think of this?<br>Are we going to see a new trend of Knowledge Distilation using diffusion models to create customizable datasets.</p><p>In the past we have seen some data augmentation using generative models like GANs so why not.</p><p><a href="https://sigmoid.social/tags/GAN" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>GAN</span></a> <a href="https://sigmoid.social/tags/GenerativeModels" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>GenerativeModels</span></a> <a href="https://sigmoid.social/tags/DataAugmentation" class="mention hashtag" rel="nofollow noopener" target="_blank">#<span>DataAugmentation</span></a></p><p>2/2</p>