Spark in me
2.32K subscribers
593 photos
40 videos
114 files
2.5K links
Lost like tears in rain. DS, ML, a bit of philosophy and math. No bs or ads.
Download Telegram
Digest 2022-03

# ML

The Gradient Update features our article - https://thegradientpub.substack.com/p/gradient-update-19-woes-of-the-irs?s=r
Constrained Reweighting for Training Deep Neural Nets with Noisy Labels - https://ai.googleblog.com/2022/02/constrained-reweighting-for-training.html
4D-Net: Learning Multi-Modal Alignment for 3D and Image Inputs in Time - https://ai.googleblog.com/2022/02/4d-net-learning-multi-modal-alignment.html
ProteInfer: deep networks for protein functional inference - https://google-research.github.io/proteinfer/
Co-training Transformer with Videos and Images Improves Action Recognition - https://ai.googleblog.com/2022/03/co-training-transformer-with-videos-and.html
Machine Learning's Most Useful Multitool: Embeddings - https://daleonai.com/embeddings-explained
Microsoft Translator enhanced with Z-code Mixture of Experts models - https://www.microsoft.com/en-us/research/blog/microsoft-translator-enhanced-with-z-code-mixture-of-experts-models/
Accelerating Ukraine Intelligence Analysis with Computer Vision on Synthetic Aperture Radar Imagery - https://bair.berkeley.edu/blog/2022/03/21/ukraine-sar-maers/
Auto-generated Summaries in Google Docs - https://ai.googleblog.com/2022/03/auto-generated-summaries-in-google-docs.html
NVIDIA Research Turns 2D Photos Into 3D Scenes in the Blink of an AI - https://blogs.nvidia.com/blog/2022/03/25/instant-nerf-research-3d-ai/?ncid=so-twit-802781-vt37#cid=gtcs22_so-twit_en-us
Instant Neural Graphics Primitives - https://nvlabs.github.io/instant-ngp/

#digest
Digest 2022-04

📌 ML

Pathways Language Model (PaLM): Scaling to 540 Billion Parameters for Breakthrough Performance - https://ai.googleblog.com/2022/04/pathways-language-model-palm-scaling-to.html

Detecting Signs of Disease from External Images of the Eye - https://ai.googleblog.com/2022/03/detecting-signs-of-disease-from.html

Reproducibility in Deep Learning and Smooth Activations - https://ai.googleblog.com/2022/04/reproducibility-in-deep-learning-and.html

VDTTS: Visually-Driven Text-To-Speech - https://ai.googleblog.com/2022/04/vdtts-visually-driven-text-to-speech.html

Discovering the systematic errors made by machine learning models - https://ai.stanford.edu/blog/domino/

Understanding BLEU Scores in Customized Machine Translation - https://blog.taus.net/understanding-bleu-scores-in-customized-machine-translation

Locked-Image Tuning: Adding Language Understanding to Image Models - https://ai.googleblog.com/2022/04/locked-image-tuning-adding-language.html

FormNet: Beyond Sequential Modeling for Form-Based Document Understanding - https://ai.googleblog.com/2022/04/formnet-beyond-sequential-modeling-for.html

Compact word vectors with Bloom embeddings - https://explosion.ai/blog/bloom-embeddings

Nobody wants your fancy algorithm - https://joemorrison.substack.com/p/nobody-wants-your-fancy-algorithm?s=r

Why Dark and Light is Complicated in Photographs - https://aaronhertzmann.com/2022/03/10/photographic-tone.html


#digest
Digest 2022-04

📌 Blogs

Horrible edge cases to consider when dealing with music - https://dustri.org/b/horrible-edge-cases-to-consider-when-dealing-with-music.html

Old bittorrent alternatives - https://habr.com/ru/post/318400/

Как врать с помощью статистики - https://habr.com/ru/post/660269/

Как мы кикшеринг взломали - https://habr.com/ru/post/660575/

TV, merchant media and the unbundling of advertising - https://www.ben-evans.com/benedictevans/2022/3/18/unbundling-advertising

Goodbye, Google Analytics - Why and How You Should Leave The
Platform - https://martinheinz.dev/blog/71

How we lost 54k GitHub stars - https://httpie.io/blog/stardust

Ускорение производительности Python в 3.11 - https://habr.com/ru/post/662087/

The Problem With Experts - https://www.strangeloopcanon.com/p/the-problem-with-experts

Netflix is not a tech company - https://www.ben-evans.com/benedictevans/2019/7/31/Netflix

Content isn't king - https://www.ben-evans.com/benedictevans/2017/7/13/content-isnt-king

#digest
Digest 2022-04

📌 Hardware

Replacing Tape with Flash - https://thessdguy.com/replacing-tape-with-flash/

HARD OR SOFT? - https://digitstodollars.com/2022/04/01/hard-or-soft/

HARD VS. SOFT – WITH MATH - https://digitstodollars.com/2022/04/07/hard-vs-soft-with-math/

Успехи импортозамещения Поднебесной: в КНР с нуля разработали игровые видеокарты и не только - https://habr.com/ru/company/selectel/blog/653807/

Объединение компьютеров через VPN и личное облако на VPS сервере - https://pc-01.tech/vpn-oblako/

Is Google Spying on your Conversations? - https://petewarden.com/2022/04/11/is-google-spying-on-your-conversations/

Иностранные хостеры с возможностью оплаты из России - https://habr.com/ru/post/657639/

WHAT IS GOING ON IN THE SEMIS SUPPLY CHAIN? - https://digitstodollars.com/2022/04/14/what-is-going-on-in-the-semis-supply-chain/

WHO SHOULD ROLL THEIR OWN CHIP? - https://digitstodollars.com/2022/04/15/who-should-roll-their-own-chip/

Перенос нейронной сети из PyTorch на Google Coral - https://habr.com/ru/company/kryptonite/blog/660505/

MAKING ALL THE CHIPS - https://digitstodollars.com/2022/04/19/making-all-the-chips/

BENCHMARKING ARM IN THE DATA CENTER - https://digitstodollars.com/2022/04/21/benchmarking-arm-in-the-data-center/

Почему GPU обманывают о своей нагрузке и как с этим бороться - https://habr.com/ru/company/yandex/blog/661989/

KEEPING UP WITH GOOGLE SEMICONDUCTOR - https://digitstodollars.com/2022/04/26/keeping-up-with-google-semiconductor/

#digest
Digest 2022-05

📌 ML

Deep Learning in Neuroimaging - https://thegradient.pub/the-role-of-deep-learning-in-understanding-neuroimaging-data/

Alpa: Automated Model-Parallel Deep Learning - https://ai.googleblog.com/2022/05/alpa-automated-model-parallel-deep.html

Rethinking Human-in-the-Loop for Artificial Augmented Intelligence - https://bair.berkeley.edu/blog/2022/05/03/human-in-the-loop/

How Should you Protect your Machine Learning Models and IP? - https://petewarden.com/2022/05/08/how-should-you-protect-your-machine-learning-models-and-ip/

Hiding a photo inside another photo - https://www.avestura.dev/blog/hide-a-photo-inside-another-photo

Unlocking Zero-Resource Machine Translation to Support New

Languages in Google Translate - https://ai.googleblog.com/2022/05/24-new-languages-google-translate.html

Baidu and Pony.ai become first robotaxi services to operate without safety drivers in Beijing - https://www.theverge.com/2022/4/30/23050493/baidu-pony-ai-first-robotaxi-services-operate-without-safety-drivers-beijing-china

Tackling multiple tasks with a single visual language model - https://www.deepmind.com/blog/tackling-multiple-tasks-with-a-single-visual-language-model

Lessons From Deploying Deep Learning To Production (it's all about feedback loops) - https://thegradient.pub/lessons-from-deploying-deep-learning-to-production/

OPT: Open Pre-trained Transformer Language Models - http://arxiv.org/abs/2205.01068
- Talk about gatekeeping: access will be granted to academic researchers; those affiliated with organizations in government, civil society, and academia; and those in industry re- search laboratories
- OPT-175B on 992 80GB A100 GPUs (1/7th the carbon footprint of GPT-3)

WHO WILL END UP HOLDING THE SEMIS BAG? - https://digitstodollars.com/2022/05/18/who-will-end-up-holding-the-semis-bag/

Image-Text Pre-training with Contrastive Captioners - https://ai.googleblog.com/2022/05/image-text-pre-training-with.html

The Future of Interactive Media — Pipelining StyleGAN3 for Production - https://medium.com/codex/the-future-of-interactive-media-pipelining-stylegan3-for-production-636c080db2f4

(De)ToxiGen: Leveraging large language models to build more robust hate speech detection tools - https://www.microsoft.com/en-us/research/blog/detoxigen-leveraging-large-language-models-to-build-more-robust-hate-speech-detection-tools/

Partnering people with large language models to find and fix bugs in NLP systems - https://www.microsoft.com/en-us/research/blog/partnering-people-with-large-language-models-to-find-and-fix-bugs-in-nlp-systems/

StarGANv2-VC: A Diverse, Unsupervised, Non-parallel Framework for Natural-Sounding Voice Conversion - https://starganv2-vc.github.io/

#digest