🎥 [ИТ-лекторий]: Яндекс.Маршрутизация: как IT-технологии улучшают логистические сервисы
👁 1 раз ⏳ 3586 сек.
Спикер: Даниил Тарарухин, руководитель группы аналитики в отделе B2BGeo геосервисов Яндекса

Описание:

Современные IT-технологии – Big Data, Machine Learning, облачные вычисления и прочие модные слова – постепенно выходят из мира Интернета и онлайн-сервисов в оффлайновый мир, стремясь освоить "классические" области экономики и сталкиваясь при этом со специфическими, порой весьма неожиданными проблемами.

Команда B2BGeo, входящая в геосервисы Яндекса (наверняка известные вам по "Яндекс.Картам" и "Навигатор

​StyleGAN2: Near-Perfect Human Face Synthesis...and More

🔗 StyleGAN2: Near-Perfect Human Face Synthesis...and More
❤️ Check out Weights & Biases here and sign up for a free demo: https://www.wandb.com/papers Their blog post on street scene segmentation is available here: https://app.wandb.ai/borisd13/semantic-segmentation/reports/Semantic-Segmentation-on-Street-Scenes--VmlldzoxMDk2OA 📝 The paper "Analyzing and Improving the Image Quality of #StyleGAN" and its source code is available here: - http://arxiv.org/abs/1912.04958 - https://github.com/NVlabs/stylegan2 - https://colab.research.google.com/drive/1ShgW6wohEFQtqs_

​Автор рассказывает, как научил искусственный интеллект играть в Space Invaders.

https://hackernoon.com/how-i-trained-an-ai-to-play-atari-space-invaders-b3e8756ef026

🔗 How I Trained an AI to Play Atari Space Invaders

​Modeling the Risk of Traffic Accidents in New York City

🔗 Modeling the Risk of Traffic Accidents in New York City
Classifying risk of traffic accidents using a convLSTM2D deep learning model

​BlenderProc
BlenderProc is a modular procedural pipeline, which helps in generating real looking images for the training of convolutional neural networks.

https://github.com/DLR-RM/BlenderProc

https://github.com/DLR-RM/BlenderProc4BOP

https://arxiv.org/abs/1911.01911v1

Наш телеграм канал - tglink.me/ai_machinelearning_big_data

🔗 DLR-RM/BlenderProc
A procedural blender pipeline to generate images for deep learning - DLR-RM/BlenderProc

​Meena: Google’s New Chatbot

🔗 Meena: Google’s New Chatbot
A more human-like and versatile chatbot

​Robots are Ready for the Real World

🔗 Robots are Ready for the Real World
The new startup, Covariant, from UC Berkeley Artificial Intelligence Research shows the robots can do it in the big game (real world).

🎥 ml5.js: Train a Neural Network with Pixels as Input
👁 2 раз ⏳ 1118 сек.
This tutorial builds on ml5.neuralNetwork() videos examining raw pixels as inputs to a neural network. This sets the stage for a discussion on convolutional neural networks.

💻Code: https://thecodingtrain.com/Courses/ml5-beginners-guide/8.1-pixels-input.html
🎥Next video: coming soon!
🎥Beginners Guide to Machine Learning: https://www.youtube.com/playlist?list=PLRqwX-V7Uu6YPSwT06y_AEYTqIwbeam3y

🔗Starting code: https://editor.p5js.org/codingtrain/sketches/ARYvi6amN
🔗Regression: https://editor.p5js.org/codingt

🎥 MIT Deep Learning Basics 2020 Introduction and Overview
👁 1 раз ⏳ 4086 сек.

​Open Source Differentiable Computer Vision Library for PyTorch

https://kornia.org

Code: https://github.com/kornia/kornia

Paper: https://arxiv.org/abs/1910.02190v2
Наш телеграм канал - tglink.me/ai_machinelearning_big_data

🔗 kornia/kornia
Open Source Differentiable Computer Vision Library for PyTorch - kornia/kornia

​Real-Time Sentiment Analytics and Visualization via ElectionTweetBoard

🔗 Real-Time Sentiment Analytics and Visualization via ElectionTweetBoard
This post will explore various forms of and considerations in data analytics, visualization, as well as Machine Learning by delving deeper…

​Топ-6 инструментов для анализа данных / data science.

https://youtu.be/23QtdnfhBRY

🔗 Top 6 Tool Types For Data Analysis / Data Science - Save hours by using the right tool
Quick Summary of YouTube Live stream Which Tools For Data Analytics / Data Science https://youtu.be/GD-JnuNS9gs R vs Python https://youtu.be/ETvvwTuiIps Data Wrangling with Excel https://www.youtube.com/playlist?list=PL8ncIDIP_e6uzAHkxmIqkHDAXkzoPWrDX

​Fast Video Object Segmentation using the Global Context Module.

http://arxiv.org/abs/2001.11243

🔗 Fast Video Object Segmentation using the Global Context Module
We developed a real-time, high-quality video object segmentation algorithm for semi-supervised video segmentation. Its performance is on par with the most accurate, time-consuming online-learning model, while its speed is similar to the fastest template-matching method which has sub-optimal accuracy. The core in achieving this is a novel global context module that reliably summarizes and propagates information through the entire video. Compared to previous approaches that only use the first, the last, or a select few frames to guide the segmentation of the current frame, the global context module allows us to use all past frames to guide the processing. Unlike the state-of-the-art space-time memory network that caches a memory at each spatiotemporal position, our global context module is a fixed-size representation that does not use more memory as more frames are processed. It is straightforward in implementation and has lower memory and computational costs than the space-time memory module. Equipped with the

​Project DeepSpeech
A TensorFlow implementation of Baidu's DeepSpeech architecture

Code: https://github.com/mozilla/DeepSpeech

Tensorflow & Pytorch: https://github.com/DemisEom/SpecAugment

SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition:

https://arxiv.org/pdf/1904.08779.pdf

Наш телеграм канал - tglink.me/ai_machinelearning_big_data

🔗 mozilla/DeepSpeech
A TensorFlow implementation of Baidu's DeepSpeech architecture - mozilla/DeepSpeech

​An Introduction to Unity ML-Agents

🔗 An Introduction to Unity ML-Agents
We’ll learn how Unity ML-Agents works and at the end of the article, you’ll train a RL agent to learn to jump over walls.

🎥 StyleGAN, Latent Space Interpolation - Week 2
👁 1 раз ⏳ 417 сек.
Learn more about machine learning for image makers by signing up at https://mailchi.mp/da905fbd76ee/machine-learning-artists

https://bustbright.square.site/
https://www.instagram.com/dvsmethid/
https://twitter.com/dvsch
https://dvschultz.github.io/design/

​GitHub - BY571/Upside-Down-Reinforcement-Learning

https://github.com/BY571/Upside-Down-Reinforcement-Learning

🔗 BY571/Upside-Down-Reinforcement-Learning
Contribute to BY571/Upside-Down-Reinforcement-Learning development by creating an account on GitHub.

​A review of machine learning for neuroscience:
https://www.mdpi.com/2076-3425/9/3/67/htm

🔗 Sketching the Power of Machine Learning to Decrypt a Neural Systems Model of Behavior
Uncovering brain-behavior mechanisms is the ultimate goal of neuroscience. A formidable amount of discoveries has been made in the past 50 years, but the very essence of brain-behavior mechanisms still escapes us. The recent exploitation of machine learning (ML) tools in neuroscience opens new avenues for illuminating these mechanisms. A key advantage of ML is to enable the treatment of large data, combing highly complex processes. This essay provides a glimpse of how ML tools could test a heuristic neural systems model of motivated behavior, the triadic neural systems model, which was designed to understand behavioral transitions in adolescence. This essay previews analytic strategies, using fictitious examples, to demonstrate the potential power of ML to decrypt the neural networks of motivated behavior, generically and across development. Of note, our intent is not to provide a tutorial for these analyses nor a pipeline. The ultimate objective is to relate, as simply as possible, how complex neuroscience c

🎥 Machine learning + neuroscience = biologically feasible computing | Benjamin Migliori | TEDxSanDiego
👁 1 раз ⏳ 720 сек.
Whether you're a human, an animal, or a machine, decisions can't be made without perception, which is how we come to understand the world around us. Machine learning will allow us to create a future in which artificial systems extend and augment our abilities, to help us create and imagine. To do that, we need to create machines that make decisions based on instinct, context, and minimal training. Combining neuroscience and machine learning, we can enter the world of biologically feasible computing. Compani

🎥 Stanford ICME Lecture on Why Deep Learning Works. Jan 2020
👁 1 раз ⏳ 4558 сек.
Random Matrix Theory (RMT) is applied to analyze the weight matrices of
Deep Neural Networks (DNNs), including production quality, pre-trained
models and smaller models trained from scratch. Empirical and theoretical
results indicate that the DNN training process itself implements a
form of self-regularization, evident in the empirical spectral density (ESD)
of DNN layer matrices. To understand this, we provide a phenomenology
to identify 5+1 Phases of Training, corresponding to increasing amounts of
i

​ML,VR & Robots (и немного облака)
Всем привет!

Хочу рассказать об очень не скучном проекте, где пересеклись робототехника, Machine Learning (а вместе это уже Robot Learning), виртуальная реальность и немного облачных технологий. И все это на самом деле имеет смысл. Ведь это и правда удобно — вселяться в робота, показывать, что ему делать, а затем обучать веса на ML сервере по сохраненным данным.

Под катом мы расскажем, как оно сейчас работает, и немного деталей про каждый из аспектов, который пришлось разрабатывать.

🔗 ML,VR & Robots (и немного облака)
Всем привет! Хочу рассказать об очень не скучном проекте, где пересеклись робототехника, Machine Learning (а вместе это уже Robot Learning), виртуальная реальн...