What is TPU ?!
(Tensor processing unit)

تصویر: http://cdn.mos.cms.futurecdn.net/95b214f2e6ed15a55df3e6a46d28f768-970-80.jpg

تی.پی.یو یا واحد پردازش تانسور نوعی "مدارهای مجتمع با کاربرد خاص" هستند که به طور خاص برای کارهای یادگیری ماشین توسعه داده شده اند.
در مقایسه با GPU ها که در سال‌های اخیر علوه بر کاربرد اصلی کارهای گرافیکی برای همین دست مسائل استفاده شده؛ TPUها برای حجم بالاتر داده ولی دقیق تر(مثلا 8 بیتی) و کاهش محاسبات طراحی شده اند.

گوگل که این TPUها را برای کارهای خودش طراحی کرده, ادعا کرده این واحدهای پردازشی تا 10 برابر سریع تر GPU ها برای کارهای #یادگیری_ماشین هستند.

"The TPU used lower precision of 8 bit and possibly lower to get similar performance to what Myriad 2 delivers today. Similar to us they optimized for use with #TensorFlow," said Dr. David Moloney, Movidius' CTO.

اطلاعات بیشتر:

Google's Big Chip Unveil For Machine Learning: Tensor Processing Unit With 10x Better Efficiency
🔗http://www.tomshardware.com/news/google-tensor-processing-unit-machine-learning,31834.html


Google's Tensor Processing Unit explained: this is what the future of computing looks like
🔗http://www.techradar.com/news/computing-components/processors/google-s-tensor-processing-unit-explained-this-is-what-the-future-of-computing-looks-like-1326915

مرتبط:

Google’s tensor processing units (TPUs) are interesting, but Nvidia is essential
🔗http://tech.firstpost.com/news-analysis/googles-tensor-processing-units-tpus-are-interesting-but-nvidia-is-essential-316577.html

Google's Tensor Processing Unit could advance Moore's Law 7 years into the future
🔗http://www.pcworld.com/article/3072256/google-io/googles-tensor-processing-unit-said-to-advance-moores-law-seven-years-into-the-future.html

#TPU #GPU #ASIC @CVISION

TensorFlow: A Flexible, Scalable & Portable System
+فیلم و اسلاید
Summary
Rajat Monga talks about why engineers at #Google built #TensorFlow, an open source software library for numerical computation using data flow graphs, and what were some of the technical challenges in building it. TensorFlow leverages a general computational model that is applicable in a wide variety of other domains, especially for performing large-scale numerical computations.

https://www.infoq.com/presentations/tensorflow

خانه ی #هوشمند مارک #زاکربرگ بنیان گذار فیس بوک که از متدهای نوین هوش مصنوعی نظیر بازشناسی شئ، بازشناسی چهره، بازشناسی گفتار، پردازش زبان‌های طبیعی و ... بهره برده است.
زاکربرگ از انگیزه ی خود برای این کار و گام های انجام کارش می‌نویسد:


https://www.facebook.com/notes/mark-zuckerberg/building-jarvis/10154361492931634/

چالش شخصی من برای سال 2016 ساخت یک هوش مصنوعی ساده برای خانه ام بوده - مثل جارویس در فیلم مرد آهنین...

Building Jarvis:
- Getting Started: Connecting the Home
- #Natural_Language
- #Vision and #Face_Recognition
- Messenger Bot
- Voice and #Speech_Recognition
- Facebook Engineering Environment

—------
Vision and Face Recognition:
About one-third of the human #brain is dedicated to vision, and there are many important #AI problems related to understanding what is happening in images and videos. These problems include #tracking (eg is Max awake and moving around in her crib?), #object_recognition (eg is that Beast or a rug in that room?), and face recognition (eg who is at the door?).
Face recognition is a particularly difficult version of object recognition because most people look relatively similar compared to telling apart two random objects — for example, a sandwich and a house. But Facebook has gotten very good at face recognition for identifying when your friends are in your photos. That expertise is also useful when your friends are at your door and your AI needs to determine whether to let them in.
To do this, I installed a few cameras at my door that can capture images from all angles. AI systems today cannot identify people from the back of their heads, so having a few angles ensures we see the person's face. I built a simple server that continuously watches the cameras and runs a two step process: first, it runs face detection to see if any person has come into view, and second, if it finds a face, then it runs face recognition to identify who the person is. Once it identifies the person, it checks a list to confirm I'm expecting that person, and if I am then it will let them in and tell me they're here.
This type of visual AI system is useful for a number of things, including knowing when Max is awake so it can start playing music or a Mandarin lesson, or solving the context problem of knowing which room in the house we're in so the AI can correctly respond to context-free requests like "turn the lights on" without providing a location. Like most aspects of this AI, vision is most useful when it informs a broader model of the world, connected with other abilities like knowing who your friends are and how to open the door when they're here. The more context the system has, the smarter is gets overall.

#mark_zuckerberg #smart_home

فیلم مرتبط:👆

برنامه زمانی سخنرانیها:
http://wss.ce.sharif.edu/#schedule

مقایسه ی فریم ورکهای یادگیری ژرف:
DL4J vs. Torch vs. Theano vs. Caffe vs. TensorFlow:

https://deeplearning4j.org/compare-dl4j-torch7-pylearn

Content:
#Theano & Ecosystem
#Torch
#TensorFlow
#Caffe
#CNTK
#DSSTNE
#Keras
#Mxnet
#Paddle
Licensing

Brain Simulator is currently being developed in C# with AI modules and architectures simulated on CUDA. We plan to make the project multi-platform (Windows, Linux, Mac, etc.), cloud-based, and cluster-based (multi-GPU/CPU).

http://www.goodai.com/brain-simulator

Brain Simulator introductory user tutorial:
https://youtu.be/4ghhVl_UJwk

demo1
https://youtu.be/froJfzmwj18
demo2:
https://youtu.be/0RZM6NNuEmc

‎ سخنرانى على اسلامى در حوزه‌ی Deep Learning
شنبه ١١ دى، ساعت ۱۶ تا ۱۸
‎محل برگزارى: آمفى‌تئاتر دانشكده‌ی مهندسى برق و کامپیوتر دانشگاه تهران

وویس سخنرانی دکتر علی اسلامی با موضوع :

Beyond Supervised Deep Learning

۸ دی، دانشگاه شریف 👇👇

#Installing #CPU and #GPU #TensorFlow on #Windows:
فیلم آموزش #نصب تنسرفلو بر روی #ویندوز
سازگار با ویندوز 10 و همچنین 7 و 8
زبان : انگلیسی

Installing CPU and GPU TensorFlow on Windows.mp4👇

انسان در واقع اشیاء را بدون ناظر یاد می‌گیرد و بعد اینکه مثلا مدتی یک شی را دید و یاد گرفت، بلافاصله پس از اینکه نام آن شی را شنید برچسب آن را نیز یاد میگیرد.
در حال حاضر بهترین مدلهای بینایی ماشین که در سالهای اخیر، خصوصا بعد از الکس‌نت سال 2012 ارائه شده اند با ناظر هستند. خیلی خوب عمل می‌کنند اما به داده ی برچسب گذاری شده ی زیادی نیاز دارند.
اگر به نحوی بتوانیم از داده های بدون برچسب استفاده کنیم و مدل را آموزش دهیم، سپس در فاز کوتاهی با داده های اندک اشیائی که مدل یاد گرفته است را به صورت با ناظرآموزش دهیم تحول بزرگی در یادگیری مدل ها ایجاد خواهد شد. در این صورت میتوان به سادگی میلیون ها ساعت ویدیو را مثلا با استفاده از یوتیوب به مدل آموزش داد و پس از آموزش مدل، شروع به آموزش نام اشیاء یادگرفته شده به مدل پرداخت روندی که در انسان هم مشاهده میشود! در واقع کودک از بدو تولد اشیاء مختلف را میبیند و آن ها را یاد میگیرد اما با یک یا چند بارشنیدن نام آن به آن دسته یا شئی که قبلا فراگرفته نام اختصاص میدهد.

The Next Frontier in AI: Unsupervised Learning
#Yann_LeCun
Director of AI Research at Facebook, Professor of Computer Science, New York University

November 18, 2016


https://www.youtube.com/watch?v=IbjF5VjniVE

Abstract
The rapid progress of #AI in the last few years are largely the result of advances in #deep_learning and neural nets, combined with the availability of large datasets and fast GPUs. We now have systems that can #recognize images with an accuracy that rivals that of humans. This will lead to revolutions in several domains such as autonomous transportation and #medical #image understanding. But all of these systems currently use #supervised learning in which the machine is trained with inputs labeled by humans. The challenge of the next several years is to let machines learn from raw, #unlabeled_data, such as #video or #text. This is known as #unsupervised learning. AI systems today do not possess "common sense", which humans and animals acquire by observing the world, acting in it, and understanding the physical constraints of it. Some of us see unsupervised learning as the key towards machines with common sense. Approaches to unsupervised learning will be reviewed. This presentation assumes some familiarity with the basic concepts of deep learning.

یان لیکان در این سخنرانی
شبکه های رقابتی مولد
یا
Generative Adversarial Networks
را مهم ترین ایده در 20 سال گذشته برای یادگیری ماشین بیان کرده است.

روشی که مدلها را قادر به یادگیری بدون ناظر می‌کند.

The major advancements in Deep Learning in 2016
🔗https://tryolabs.com/blog/2016/12/06/major-advancements-deep-learning-2016/

Generative Adversarial Nets
https://arxiv.org/pdf/1406.2661v1.pdf

این روش برای مسائل با تعداد کم و ناکافی داده ی با برچسب نیز مناسب است.
#autoencoder #unsupervised #unsupervised_learning #Generative #Generative_Models

✏️Title:
#Unsupervised #Representation Learning with #Deep #Convolutional #Generative #Adversarial Networks
✏️abstract:
In recent years, supervised learning with convolutional networks (CNNs) has seen huge adoption in computer vision applications. Comparatively, unsupervised learning with CNNs has received less attention. In this work we hope to help bridge the gap between the success of CNNs for supervised learning and unsupervised learning. We introduce a class of CNNs called deep convolutional generative adversarial networks (DCGANs), that have certain architectural constraints, and demonstrate that they are a strong candidate for unsupervised learning. Training on various image datasets, we show convincing evidence that our deep convolutional adversarial pair learns a hierarchy of representations from object parts to scenes in both the generator and discriminator. Additionally, we use the learned features for novel tasks - demonstrating their applicability as general image representations.

🔗https://arxiv.org/pdf/1511.06434v2.pdf

"Under review as a conference paper at ICLR 2016"

The talks at the #Deep_Learning #School on September 24/25, 2016
http://www.bayareadlschool.org


Full Day Live Streams:
Day 1: https://youtu.be/eyovmAtoUx0
Day 2: https://youtu.be/9dXiAecyJrY

دو ویدیوی بالا تفکیک نشده اند و هر کدام حدود 10 ساعت سخرانی است
در زیر هر سخرانی جدا شده است:

1. Foundations of Deep Learning (Hugo Larochelle, Twitter) - https://youtu.be/zij_FTbJHsk
2. Deep Learning for Computer Vision (Andrej Karpathy, OpenAI) - https://youtu.be/u6aEYuemt0M
3. Deep Learning for Natural Language Processing (Richard Socher, Salesforce) - https://youtu.be/oGk1v1jQITw
4. TensorFlow Tutorial (Sherry Moore, Google Brain) - https://youtu.be/Ejec3ID_h0w
5. Foundations of Unsupervised Deep Learning (Ruslan Salakhutdinov, CMU) - https://youtu.be/rK6bchqeaN8
6. Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I
7. Deep Reinforcement Learning (John Schulman, OpenAI) - https://youtu.be/PtAIh9KSnjo
8. Theano Tutorial (Pascal Lamblin, MILA) - https://youtu.be/OU8I1oJ9HhI
9. Deep Learning for Speech Recognition (Adam Coates, Baidu) - https://youtu.be/g-sndkf7mCs
10. Torch Tutorial (Alex Wiltschko, Twitter) - https://youtu.be/L1sHcj3qDNc
11. Sequence to Sequence Deep Learning (Quoc Le, Google) - https://youtu.be/G5RY_SUJih4
12. Foundations and Challenges of Deep Learning (Yoshua Bengio) - https://youtu.be/11rsu_WwZTc

Learning from Simulated and Unsupervised Images through Adversarial Training (Apple Inc.)

مقاله‌ی جالب کمپانی اپل!
( Submitted for review to a conference on Nov 15, 2016)

✏️Title:
Learning from Simulated and Unsupervised Images through Adversarial Training

✏️abstract:
With recent progress in graphics, it has become more tractable to train models on #synthetic images, potentially avoiding the need for expensive annotations. However, learning from synthetic images may not achieve the desired performance due to a gap between synthetic and real image distributions. To reduce this gap, we propose Simulated+Unsupervised (S+U) learning, where the task is to learn a model to improve the realism of a simulator's output using #unlabeled real data, while preserving the annotation information from the simulator. We develop a method for S+U learning that uses an #adversarial network similar to #Generative Adversarial Networks (#GANs), but with synthetic images as inputs instead of random vectors. We make several key modifications to the standard GAN algorithm to preserve annotations, avoid artifacts and stabilize training: (i) a 'self-regularization' term, (ii) a local adversarial loss, and (iii) updating the discriminator using a history of refined images. We show that this enables generation of highly realistic images, which we demonstrate both qualitatively and with a user study. We quantitatively evaluate the generated images by training models for gaze estimation and hand pose estimation. We show a significant improvement over using synthetic images, and achieve state-of-the-art results on the MPIIGaze dataset without any labeled real data.

🔗https://arxiv.org/abs/1612.07828v1
🔗https://arxiv.org/pdf/1612.07828v1.pdf

#unlabeled_data #unsupervised #unsupervised_learning #Generative #Generative_Models