Google engineers offered 28 actionable tests for #machinelearning systems. 👇

Introducing 👉 The ML Test Score: A Rubric for ML Production Readiness and Technical Debt Reduction (2017). 👈

If #ml #training is like compilation, then ML testing shall be applied to both #data and code.

7 model tests

1⃣ 👉 Review model specs and version-control it. It makes training auditable and improve reproducibility.

2⃣ 👉 Ensure model loss is correlated with user engagement.

3⃣ 👉 Tune all hyperparameters. Grid search, Bayesian method whatever you use, tune all of them.

4⃣ 👉 Measure the impact of model staleness. The age-versus-quality curve shows what amount of staleness is tolerable.

5⃣ 👉 Test against a simpler model regularly to confirm the benefit more sophisticated techniques.

6⃣ 👉 Check the model quality is good across different data segment, e.g. user countries, movie genre etc.

7⃣ 👉 Test model inclusion by checking against the protected dimensions or enrich under-represented categories.

7 data tests

1⃣ 👉 Capture feature expectations in schema using statistics from data + domain knowledge + expectations.

2⃣ 👉 Use beneficial features only, e.g. training a set of models each with one feature removed.

3⃣ 👉 Avoid costly features. Cost includes running time, RAM as well as upstream work and instability. 

4⃣ 👉 Adhere to feature requirements. If certain features can’t be used, enforce it programmatically.

5⃣ 👉 Set privacy controls. Budget enough time for new feature that depends on sensitive data.

6⃣ 👉 Add new features quickly. If conflicting with 5⃣ , privacy goes first.

7⃣ 👉 Test code for all input features. Bugs do exist in feature creation code.

See 7 Infrastructure & 7 monitoring tests in paper. 👇

They interviewed 36 teams across Google and found

👉 Using a checklist helps avoid mistakes (like a surgeon would do).

👉 Data dependencies leads to outsourcing responsibility. Other teams’ validation may not validate your use case.

👉 A good framework promotes integration test which is not well adopted.

👉 Assess the assessment to better assess your system.
https://research.google.com/pubs/archive/aad9f93b86b7addfea4c419b9100c6cdd26cacea.pdf

✔️ چرا اپل تنها یک شرکت تکنولوژی نیست؟

#business
#کسب‌وکار

@Theeconomicspapers

چالشها و چشم اندازهای هوش مصنوعی و یادگیری ماشین
❇️ @ai_python

#هینتون : یک مثال عملی برای RNN
@ai_python

Why google Bert was ahead of it's time?

Because it was a masked language model even before COVID
😂😂😂😂😂

@ai_python

10 Free Resources To Learn PyTorch In 2022

List of curated PyTorch resources:

1) PyTorch Official Tutorials.

2) Intro to Deep Learning with PyTorch by Facebook AI.

3) PyTorch Fundamentals By Microsoft

4) PyTorch - Python Deep Learning Neural Network API by Deeplizard.

5) Deep Neural Networks with PyTorch by Joseph Santarcangelo

6) PyTorch Basics for Machine Learning by IBM

7) Deep Learning with Python and PyTorch

8) Pytorch - Deep learning with Python by Harrison Kinsley, Sentdex

9) Make Your First GAN Using PyTorch

10) PyTorch Tutorials By Morvan Zhou

Bonus:
Deep Learning with PyTorch book📚

#فیلم #کلاس_آموزشی #هوش_مصنوعی #الگوریتمها

❇️ @AI_Python

کانال ما حامی بنیاد نیکوکاری نسیم مهربانی هست . علاقمندان به شرکت در این مهربانی در حدی که علاقه دارند کمک کنید

تولید شده توسط هوش مصنوعی

عکس اول (بالا سمت چپ) چیزی هست که با توضیحات من (مهندسی) توسط stable diffusion تولید شده (اصن جذاب نیست)
برای همین از یک دوست طراح خوساتم کهکشان رو با چشم طراحی برام توصیف کنه و چندتا جمله بهم بده (درک اینکه چی میگه و تو ذهنش چی هست برام سخت بود) اما بعد از تولید stable diffusion و انتخاب بین عکسا (برای هر متن ۵ تا عسک تولید شد) قشنگ درک کردم توضیحاتش رو که خودش یک قدم بزرگی هست بنظرم.

خواستم بگم به جزئیات عکس‌ها نگاه کنید و مدل‌های هوش مصنوعیتون رو به طراح‌ها بسپارید.

پ.ن : عکس‌ها رو نذاشتم چون watermark - safety checker رو توی سورس کد خاموش کردم :
بخاطر اینکه بفهمم روی چه دیتایی آموزش دیده (بله امکانش هست)

اگر میخواید توسط بستر اسپیس هاگینگ فیس امتحان کنید (که البته اینقدر بهش هجوم آوردن سخت جواب میده)
https://huggingface.co/spaces/stabilityai/stable-diffusion

اگر هم می‌خواید خودتون کد رو ران کنید (اگر خودتون GPU دارید یا حتی با کولب هم میشه ران گرفت):
https://huggingface.co/CompVis/stable-diffusion-v1-4

کیفیت تصاویری خروجیش خیلی عالیه که نشون میده ایده ای Latent Diffusion خوب کار میکنه اما ارتباط اش با متن از اونی که توی مدل Imagen دیده بودیم خیلی کمتره. این مدل حجم Text Encoder اش تقریبا 10 برابر کوچیک تر از Imagen هست. مقاله Imagen اشاره کرده بود که حجم مدل Text Encoder خیلی مهمه و هر چقدر بزرگ تر باشه تصاویر خیلی مرتبط تری تولید می شه.
اینجا می تونید چند تا از نمونه های خروجی Imagen رو ببینید و مقایسه کنید:‌
https://imagen.research.google/

@cvisio

اگه می خواید به صورت تخصصی تر بررسی کنید از مقاله زیر استفاده کنید .
https://arxiv.org/abs/2112.10752
#مقاله

✔️ نقشه راه برای تبدیل شدن به یک Smart Contract Auditor
@ai_python
@soliditylang_fa
@theeconomicspapers

Learning Affordance Grounding from Exocentric Images

Github: https://github.com/lhc1224/Cross-View-AG

Paper: https://arxiv.org/abs/2208.13196v1

Dataset: https://paperswithcode.com/dataset/3d-affordancenet

Independent Encoder for Deep Hierarchical Unsupervised Image-to-Image Translation

Github: https://github.com/Elvinky/IEGAN

Paper: https://arxiv.org/abs/2107.02494

Datasets: https://github.com/Elvinky/IEGAN/tree/main/dataset

Image-to-Image Translation: https://paperswithcode.com/task/image-to-image-translation

اینجا کلی سرویس ابری رایگان هست که می‌تونید پروژه‌های آزمایشی خودتونو روشون پیاده کنید:
https://github.com/ripienaar/free-for-dev

#منابع

❇️ @AI_Python

مصاحبه‌ جذابی از گروه Fully connected با دکتر Vahid Behzadan و با موضوع امنیت سیستم‌های هوشمند انجام شده است. از طریق لینک زیر می‌تونید به ویدیو دسترسی داشته باشید.


https://youtu.be/8LGH9lfBDXw

#فیلم

❇️ @AI_Python

A Deep Learning Road Map And Where To Start

https://medium.com/@ArianAmani/the-deep-learning-road-map-that-i-took-c29120b0f5e2

Text-Free Learning of a Natural Language Interface for Pretrained Face Generators


pip install git+https://github.com/openai/CLIP.git

Github: https://github.com/duxiaodan/fast_text2stylegan

Paper: https://arxiv.org/abs/2209.03177v1

Dataset: https://paperswithcode.com/dataset/ffhq

100s 𝗼𝗳 𝗡𝗟𝗣 𝗣𝗮𝗽𝗲𝗿𝘀 𝘄𝗶𝘁𝗵 𝗖𝗼𝗱𝗲! 💡

I found this amazing website, it can help you keep track of recent advancements in NLP, by offering an updated list of the latest NLP research papers.

Also it provides a link to the Github repos, making it a valuable resource for ML researchers and practitioners. You can check the following link:
https://index.quantumstat.com/

#منابع #پردازش_زبان_طبیعی #مقاله #کتاب

❇️ @AI_Python

GNNs Learn To Smell & Awesome NeurReps

1) Back in 2019, Google AI started a project on learning representations of smells. From basic chemistry we know that aromaticity depends on the molecular structure, e.g., cyclic compounds. In fact, the whole group of ”aromatic hydrocarbons” was named aromatic because they actually has some smell (compared to many non-organic molecules). If we have a molecular structure, we can employ a GNN on top of it and learn some representations - that is a tl;dr of smell representation learning with GNNs.

Recently, Google AI released a new blogpost describing the next phase of the project - the Principal Odor Map that is able to group molecules in “odor clusters”. The authors conducted 3 cool experiments: classifying 400 new molecules never smelled before and comparison to the averaged rating of a group of human panelists; linking odor quality to fundamental biology; and probing aromatic molecules on their mosquito repelling qualities. The GNN-based model shows very good results - now we can finally claim that GNNs can smell! Looking forward for GNNs transforming the perfume industry 📈

2) The NeurReps commnuity (Symmetry and Geometry in Neural Representations) is curating the Awesome List of resources and research related to the geometry of representations in the brain, deep networks, and beyond. A great resource for Neuroscience and Geometric DL folks to learn about the adjacent field!