​Illustrated: Efficient Neural Architecture Search

🔗 Illustrated: Efficient Neural Architecture Search
Macro and micro search strategies in ENAS

🎥 Really Quick Questions - Nvidia Research
👁 21 раз ⏳ 490 сек.
Nvidia is the inventor of the GPU! This tech company based in Silicon Valley has played a huge role in the deep learning revolution (which has relied primarily on GPUs for computing), and its transformed many industries. In this video, I interview Bryan Catanzaro, the Vice President of Applied Deep Learning Research at NVIDIA. Bryan got his PhD in AI from Berkeley, invented a language called Copperhead, and is an expert in parallel programming theory. Nvidia invited me to their annual conference, so I took

​Making Art With Your Webcam

🔗 Making Art With Your Webcam
An implementation and explanation of Fast Style Transfer with a precursor in Basic Style Transfer

​Pruned Cross Validation for hyperparameter optimization

🔗 Pruned Cross Validation for hyperparameter optimization
The technique’s motivation, design, and implementation

Курс «NLP with Deep Learning»

Наш телеграм канал - tglink.me/ai_machinelearning_big_data
1. Natural Language Processing with Deep Learning
2. Word Vector Representations: word2vec
3. GloVe: Global Vectors for Word Representation
4. Word Window Classification and Neural Networks
5. Backpropagation and Project Advice
6. Dependency Parsing
7. Introduction to TensorFlow
8. Recurrent Neural Networks and Language Models
9. Machine Translation and Advanced Recurrent LSTMs and GRUs

🎥 Lecture 1 | Natural Language Processing with Deep Learning
👁 877 раз ⏳ 4301 сек.
Lecture 1 introduces the concept of Natural Language Processing (NLP) and the problems NLP faces today. The concept of representing words as numeri...

🎥 Lecture 2 | Word Vector Representations: word2vec
👁 115 раз ⏳ 4697 сек.
Lecture 2 continues the discussion on the concept of representing words as numeric vectors and popular approaches to designing word vectors.

Key ...

🎥 Lecture 3 | GloVe: Global Vectors for Word Representation
👁 132 раз ⏳ 4720 сек.
Lecture 3 introduces the GloVe model for training word vectors. Then it extends our discussion of word vectors (interchangeably called word embeddi...

🎥 Lecture 4: Word Window Classification and Neural Networks
👁 72 раз ⏳ 4603 сек.
Lecture 4 introduces single and multilayer neural networks, and how they can be used for classification purposes.

Key phrases: Neural networks. Fo...

🎥 Lecture 5: Backpropagation and Project Advice
👁 31 раз ⏳ 4700 сек.
Lecture 5 discusses how neural networks can be trained using a distributed gradient descent technique known as back propagation.

Key phrases: Neur...

🎥 Lecture 6: Dependency Parsing
👁 50 раз ⏳ 4987 сек.
Lecture 6 covers dependency parsing which is the task of analyzing the syntactic dependency structure of a given input sentence S. The output of a ...

🎥 Lecture 7: Introduction to TensorFlow
👁 72 раз ⏳ 4354 сек.
Lecture 7 covers Tensorflow. TensorFlow is an open source software library for numerical computation using data flow graphs. It was originally deve...

🎥 Lecture 8: Recurrent Neural Networks and Language Models
👁 36 раз ⏳ 4683 сек.
Lecture 8 covers traditional language models, RNNs, and RNN language models. Also reviewed are important training problems and tricks, RNNs for oth...

​В коре вашего мозга 17 млрд компьютеров

В мозг поступает информация из внешнего мира, его нейроны получают данные на входе, производят обработку и выдают некий результат. Это может быть мысль (хочу карри на ужин), действие (сделать карри), изменение настроения (ура, карри!). Что бы ни получилось на выходе, это «что-то» является преобразованием данных со входа (меню) в результат на выходе («куриный дхансак, пожалуйста»). И если представлять мозг как преобразователь с входом в выходом, то неизбежна аналогия с компьютером.

Для одних это просто полезный риторический приём, для других — серьёзная идея. Но мозг — это не компьютер. Компьютером является каждый нейрон. В коре головного мозга 17 миллиардов компьютеров.
https://habr.com/ru/post/445420/

🔗 В коре вашего мозга 17 млрд компьютеров
Нейросеть нейросетей Изображение brentsview под лицензией CC BY-NC 2.0 В мозг поступает информация из внешнего мира, его нейроны получают данные на входе, про...

​Tuning a Multi-Task Fate Grand Order Trained Pytorch Network

🔗 Tuning a Multi-Task Fate Grand Order Trained Pytorch Network
building and tuning pytorch based multi-task networks

Watch Me Build an AI Startup
https://www.youtube.com/watch?v=NzmoPqte4V4

🎥 Watch Me Build an AI Startup
👁 6 раз ⏳ 2374 сек.
I'm going to build a medical imaging classification app called SmartMedScan! The potential customers for this app are medical professionals that need to scale and improve the accuracy of their diagnoses using AI. From ideation, to logo design, to integrating features like payments and AI into a single app, I'll show you my 10 step process. I hope that by seeing my thought process and getting familiar with the sequence of steps I'll demonstrate,, you too will be as inspired as I am to use this technology to

​Python Tutorial: A Name Lookup Table for Fuzzy Name Data Sets

🔗 Python Tutorial: A Name Lookup Table for Fuzzy Name Data Sets
Increase accuracy in person name matching by using name component combinations

🎥 This AI is The “Photoshop” of Human Faces
👁 1 раз ⏳ 192 сек.
❤️ Pick up cool perks on our Patreon page: https://www.patreon.com/TwoMinutePapers

📝 The paper "SC-FEGAN: Face Editing Generative Adversarial Network with User's Sketch and Color" is available here:
https://arxiv.org/abs/1902.06838
https://github.com/JoYoungjoo/SC-FEGAN

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
313V, Alex Haro, Andrew Melnychuk, Angelos Evripiotis, Anthony Vdovitchenko, Brian Gilman, Christian Ahlin, Christoph Jadanowski, Claudio Fernand

​Geo Experiments, The Perfect Complement to A/B Testing

🔗 Geo Experiments, The Perfect Complement to A/B Testing
What Is It and How Will It Help You In Marketing?

🎥 Getting Started with TensorFlow and Deep Learning | SciPy 2018 Tutorial | Josh Gordon
👁 1 раз ⏳ 9679 сек.
A friendly introduction to Deep Learning, taught at the beginner level. We’ll work through introductory exercises across several domains - including computer vision, natural language processing, and structured data classification. We’ll introduce TensorFlow - the world’s most popular open source machine learning library - preview the latest APIs (including Eager Execution), discuss best practices, and point you to recommended educational resources you can use to learn more.

Tutorial instructions may be fou

​Другой Github 2: машинное обучение, датасеты и Jupyter Notebooks
Несмотря на то, что в интернете существует множество источников свободного программного обеспечения для машинного обучения, Github остается важным центром обмена информацией для всех типов инструментов с открытым исходным кодом, используемых в сообществе специалистов по машинному обучению и анализу данных.

В этой подборке собраны репозитории по машинному обучению, датасетам и Jupyter Notebooks, ранжированные по количеству звезд. В предыдущей части мы рассказывали о популярных репозиториях для изучения работ по визуализации данных и глубокому обучению.
https://habr.com/ru/company/mailru/blog/445530/

🔗 Другой Github 2: машинное обучение, датасеты и Jupyter Notebooks
Несмотря на то, что в интернете существует множество источников свободного программного обеспечения для машинного обучения, Github остается важным центром обме...

​Neural Quantum States — представление волновой функции нейронной сетью

В этой статье мы рассмотрим необычное применение нейронных сетей в целом и ограниченных машин Больцмана в частности для решения двух сложных задач квантовой механики — поиска энергии основного состояния и аппроксимации волновой функции системы многих тел.
https://habr.com/ru/company/raiffeisenbank/blog/445516/

🔗 Neural Quantum States — представление волновой функции нейронной сетью
В этой статье мы рассмотрим необычное применение нейронных сетей в целом и ограниченных машин Больцмана в частности для решения двух сложных задач квантовой меха...

​Unifying Physics and Deep Learning with TossingBot

🔗 Unifying Physics and Deep Learning with TossingBot
Posted by Andy Zeng, Student Researcher, Robotics at Google Though considerable progress has been made in enabling robots to grasp objec...

​Word Vectors and Lexical Semantics (Part 1)

🔗 Word Vectors and Lexical Semantics (Part 1)
The following are my personal notes based on the Deep NLP course by Oxford University held in 2017. The material is available at [1].

​Training deep neural networks on a GPU with PyTorch

🔗 Training deep neural networks on a GPU with PyTorch
Part 4 of “PyTorch: Zero to GANs”

​How to correctly select a sample from a huge dataset in machine learning

🔗 How to correctly select a sample from a huge dataset in machine learning
Choosing a small, representative dataset from a large population can improve model training reliability

​Monte Carlo Integration is Magic

🔗 Monte Carlo Integration is Magic
How to compute an integral in 3 lines of code

​Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition
https://www.shortscience.org/paper?bibtexKey=journals/corr/1406.4729

🔗 Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition - ShortScience.org
Spatial Pyramid Pooling (SPP) is a technique which allows Convolutional Neural Networks (CNNs) to use input images of any size, not only $224\text{px} \times 224\text{px}$ as most architectures do. (However, there is a lower bound for the size of the input image). ## Idea * Convolutional layers operate on any size, but fully connected layers need fixed-size inputs * Solution: * Add a new SPP layer on top of the last convolutional layer, before the fully connected layer * Use an approach similar to bag of words (BoW), but maintain the spatial information. The BoW approach is used for text classification, where the order of the words is discarded and only the number of occurences is kept. * The SPP layer operates on each feature map independently. * The output of the SPP layer is of dimension $k \cdot M$, where $k$ is the number of feature maps the SPP layer got as input and $M$ is the number of bins. Example: We could use spatial pyramid pooling with