​Машинное обучение в микрофинансах: строим скоринговую модель для клиентов с пустой кредитной историей

Наш телеграм канал - tglink.me/ai_machinelearning_big_data
Нет кредитной истории — не дают кредиты, не дают кредиты — нет кредитной истории. Замкнутый круг какой-то. Что делать? Давайте разбираться.

Привет! Меня зовут Марк, я data scientist в компании Devim. Недавно мы запустили модель для скоринга заемщиков МФК “До Зарплаты”, у которых отсутствует кредитная история. Хочу поделиться опытом поиска данных, особенностями конструирования и интерпретации признаков.

https://habr.com/ru/post/454574/

🔗 Машинное обучение в микрофинансах: строим скоринговую модель для клиентов с пустой кредитной историе
Нет кредитной истории — не дают кредиты, не дают кредиты — нет кредитной истории. Замкнутый круг какой-то. Что делать? Давайте разбираться. Привет! Меня зовут М...

​An Explicitly Relational Neural Network Architecture

🔗 An Explicitly Relational Neural Network Architecture
With a view to bridging the gap between deep learning and symbolic AI, we present a novel end-to-end neural network architecture that learns to form propositional representations with an explicitly relational structure from raw pixel data. In order to evaluate and analyse the architecture, we introduce a family of simple visual relational reasoning tasks of varying complexity. We show that the proposed architecture, when pre-trained on a curriculum of such tasks, learns to generate reusable representations that better facilitate subsequent learning on previously unseen tasks when compared to a number of baseline architectures. The workings of a successfully trained model are visualised to shed some light on how the architecture functions.

​Anomaly detection in time series with Prophet library

🔗 Anomaly detection in time series with Prophet library
Find outliers in time series and plot in few lines of code

​Wasserstein Style Transfer
arxiv.org/abs/1905.12828

🔗 Wasserstein Style Transfer
We propose Gaussian optimal transport for Image style transfer in an Encoder/Decoder framework. Optimal transport for Gaussian measures has closed forms Monge mappings from source to target distributions. Moreover interpolates between a content and a style image can be seen as geodesics in the Wasserstein Geometry. Using this insight, we show how to mix different target styles , using Wasserstein barycenter of Gaussian measures. Since Gaussians are closed under Wasserstein barycenter, this allows us a simple style transfer and style mixing and interpolation. Moreover we show how mixing different styles can be achieved using other geodesic metrics between gaussians such as the Fisher Rao metric, while the transport of the content to the new interpolate style is still performed with Gaussian OT maps. Our simple methodology allows to generate new stylized content interpolating between many artistic styles. The metric used in the interpolation results in different stylizations.

🎥 #DataScience, Леонид Кулигин, Тренировка моделей машинного обучения на больших объемах данных
👁 1 раз ⏳ 2459 сек.
Леонид Кулигин
Google

Тренировка моделей машинного обучения на больших объемах данных

Объём доступных данных, как и сложность моделей машинного обучения, растёт экспоненциально. Мы поговорим про то, как работает распределённое обучение с TensorFlow, как устроен процессинг данных и какие средства доступны для профайлинга тренировки.

https://codefest.ru

🎥 TensorFlow Dev Summit 2019 Summary + Deep Diving into GANs
👁 1 раз ⏳ 6503 сек.
During the event we will watch to a couple of key talks from the recent TensorFlow Summit showing the new features of the release 2 and in particular the new tf.function and AutoGraph allowing to write graph code using natural Python syntax.

After the vew party we will feature the talk of Paolo Galeone, Google Development Expert in ML, on how to build Generative Adversarial Networks using TensorFlow 2.0.

Deep Diving into GANs: from theory to production

Tensorflow 2.0 will be a major milestone for the mos

🎥 DeepMind Made a Math Test For Neural Networks
👁 36 раз ⏳ 301 сек.
📝 The paper "Analysing Mathematical Reasoning Abilities of Neural Models" is available here:
https://arxiv.org/abs/1904.01557

❤️ Pick up cool perks on our Patreon page: https://www.patreon.com/TwoMinutePapers

🙏 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, Bruno Brito, Bryan Learn, Christian Ahlin, Christoph Jadanowski, Claudio Fernandes, Daniel Hasegan, Dennis Abts, Eri

🎥 Введение в глубокое обучение: регрессия, классификация, свёрточные нейронные сети. 2019.
👁 1 раз ⏳ 209 сек.
⚡⚡⚡ Подводим итоги прошедших практических занятий по разработке нейронных сетей.

1. Что такое регрессия и какие задачи относятся к этому классу задач?
2. Что такое классификация и какие задачи относятся к этому классу задач?
3. Сравнение задач регрессий и классификаций - входные и выходные данные, функции оптимизации, функции активации, функции потерь.
4. Свёрточные нейронные сети и определение кошек / собак на изображениях.

Статья на Хабре:
https://habr.com/ru/post/454034/

Google CoLab:
https://colab.re

🎥 Обзор конференции AAMAS 2019
👁 1 раз ⏳ 2746 сек.
В связи с быстрым развитием в области обучения с подкреплением и искусственного интеллекта все сложнее уследить за новыми статьям. Недавно прошла одна из самых важных конференции по обучению с подкреплением — Autonomous Agents and Multiagent Systems (AAMAS). В данном докладе будут упомянуты все самый интересные идеи и статьи, которые были озвучены на конференции.

На семинаре будут рассматриваться статьи из deepmind: о новых окружениях для мультиагентных систем, также будет рассмотрены способы обучения аге

​Independent Component Analysis based on multiple data-weighting. arxiv.org/abs/1906.00028

🔗 Independent Component Analysis based on multiple data-weighting
Independent Component Analysis (ICA) - one of the basic tools in data analysis - aims to find a coordinate system in which the components of the data are independent. In this paper we present Multiple-weighted Independent Component Analysis (MWeICA) algorithm, a new ICA method which is based on approximate diagonalization of weighted covariance matrices. Our idea is based on theoretical result, which says that linear independence of weighted data (for gaussian weights) guarantees independence. Experiments show that MWeICA achieves better results to most state-of-the-art ICA methods, with similar computational time.

Introducing TensorNetwork, an Open Source Library for Efficient Tensor Calculations

http://ai.googleblog.com/2019/06/introducing-tensornetwork-open-source.html

​Moneyball — Linear Regression

🔗 Moneyball — Linear Regression
Using Linear Regression in Python to predict baseball season performance.

🎥 Childhood's End: Maturation of Deep Speech and Common Voice
👁 2 раз ⏳ 1048 сек.
#reworkDL

This presentation took place at the Deep Learning Summit, Boston on the 23 & 24 May. This presentation was given by Kelly Davis, Machine Learning Researcher at Mozilla.

More presentations and interviews from the Deep Learning Summit can be seen here on the Video Hub: http://videos.re-work.co/events/59-deep-learning-summit-boston-2019

We’ll talk about the blossoming of Deep Speech, an open deep learning based speech-to-text engine, and Common Voice, an open crowd-sourced speech corpora. We will

​DeepMind Made a Math Test For Neural Networks

https://arxiv.org/abs/1904.01557

🔗 Analysing Mathematical Reasoning Abilities of Neural Models
Mathematical reasoning---a core ability within human intelligence---presents some unique challenges as a domain: we do not come to understand and solve mathematical problems primarily on the back of experience and evidence, but on the basis of inferring, learning, and exploiting laws, axioms, and symbol manipulation rules. In this paper, we present a new challenge for the evaluation (and eventually the design) of neural architectures and similar system, developing a task suite of mathematics problems involving sequential questions and answers in a free-form textual input/output format. The structured nature of the mathematics domain, covering arithmetic, algebra, probability and calculus, enables the construction of training and test splits designed to clearly illuminate the capabilities and failure-modes of different architectures, as well as evaluate their ability to compose and relate knowledge and learned processes. Having described the data generation process and its potential future expansions, we condu

​Advanced Topics in Deep Convolutional Neural Networks

🔗 Advanced Topics in Deep Convolutional Neural Networks
Residual networks, saliency maps, dilated convolutions, and more.

​Private AI — Federated Learning with PySyft and PyTorch

🔗 Private AI — Federated Learning with PySyft and PyTorch
An application to SMS spam detection with a GRU model

🎥 4 Ways to Use Machine Learning for Mobile
👁 1 раз ⏳ 3020 сек.
Learn more about AWS Startups at – https://amzn.to/2WG04um 
Machine Learning at the edge is about making your apps smarter with real-time object and face recognition, offline predictions, and protecting user privacy. In this session, we'll explore the four ways to handle prediction or inference and decision making in modern apps. Demystify deep learning and easily call managed ML services, build, train, and/or deploy ML models to mobile and IoT devices.

🎥 An Introduction to Deep Learning
👁 1 раз ⏳ 2627 сек.
Learn more about AWS Startups at – https://amzn.to/2WhgEwo 
Getting started with deep learning can feel really intimidating. In this session we'll dive right in to explaining the basic concepts of deep learning with barely any jargon and a small amount of arithmetic. We'll see how easy it is to take an existing pre-trained general-purpose image classification model from the cloud and re-train it to identify objects that we want the computer to recognize. We'll show how to do all of this with python, using a

🎥 My Experience With Deep Learning and TensorFlow | Nathan Hadjiyski | TEDxYouth@AnnArbor
👁 1 раз ⏳ 674 сек.
Nathan Hadjiyski is a 10th grader at Pioneer High School. He has been interested in science from young age and his passion for it continues to grow. He is trying to benefit society with his cancer research, and he hopes it could ultimately help patients get a more accurate diagnosis and as a result more adequate treatment. Nathan Hadjiyski’s presentation is about his experience with Deep Learning and Tensor Flow applied to kidney cancer diagnosis. Nathan Hadjiyski is a 10th grader at Pioneer High School. He

🎥 Обзор ICLR 2019
👁 1 раз ⏳ 3171 сек.
Девятого мая закончилась очередная International Conference on Learning Representations (ICLR). Мы сделаем обзор публикаций с ICLR, которые нам показались наиболее интересными.

Докладчик: Рауф Курбанов.

Ссылка на слайды: https://research.jetbrains.org/files/material/5cf7a348264f3.pdf