Top #Kaggle masters are releasing image augmentation library v0.1.1 with an extended bounding box support.

Guthub: https://github.com/albu/albumentations/releases/tag/v0.1.1

#cv #dl #dataaugmentation

Astrologers proclaimed month of #dataaugmentation since #GoogleAI released AutoAugment library — reinforcement learning algorithm which increases both the amount and diversity of existing data by finding optimal image augmentation policies.


Link: https://ai.googleblog.com/2018/06/improving-deep-learning-performance.html
Arxiv: https://arxiv.org/abs/1805.09501

Can Global Semantic Context Improve Neural Language Models?

New article by #Apple Frameworks Natural Language Processing Team

https://machinelearning.apple.com/2018/09/27/can-global-semantic-context-improve-neural-language-models.html

#nlp

Large scale GAN training for
high fidelity natural image synthesis

Link: https://openreview.net/pdf?id=B1xsqj09Fm
Samples: https://drive.google.com/drive/folders/1lWC6XEPD0LT5KUnPXeve_kWeY-FxH002

#gan #dl #cv

🎂🎉New Release - #Matplotlib 3.0.0. Supports Python 3. Highlights include:

GUI backend is selected at run-time based on what toolkits are installed;
New cyclic color map *twilight*;
Improvements to automatic layout of titles, ticks & GridSpec.

mail thread: https://mail.python.org/pipermail/matplotlib-announce/2018-September/000027.html
official site: https://matplotlib.org/users/whats_new.html
installation: pip install -U matplotlib

#visualization #dataviz

Introduction for machine learning for coders

Fast.ai launching new course for coders, having at least 1 year of experience. This is a practical-oriented course covering wide range of areas: classical Machine Learning with Random Forest and Gradient Decent, Regularization, NLP, Embeddings,

Link: https://www.fast.ai/2018/09/26/ml-launch/
Course itself: https://course.fast.ai/ml

#fastai #novice #entrylevel

AR Trajectory prediction

This is a real prototype with all the 30 lines of code to reimplement it.

Source: https://www.3delement.com/?p=610

#AR

Feature selection — Correlation and P-value

Basic article, explaining key concepts: #correlation and #p_value with code example.

Link: https://towardsdatascience.com/feature-selection-correlation-and-p-value-da8921bfb3cf

#novice

What does your Spotify music sound like? Data Science with Spotify (Part 1)

Example of a good approach to the research. Though, as was noted, there is no data for the reproducibility, author can provide data and sample code in the future.

Link: https://towardsdatascience.com/data-science-and-machine-learning-with-spotify-841225bfb5d0

#spotify

Reproducing Imagenet in 18 minutes

The code to reproduce #ImageNet in 18 minutes is posted in the GitHub repo. It actually becomes «Imagenet in 12 minutes» if using 74.9% top1, used in Chainer's "Imagenet in 15" paper, last few bits are the hardest.

Link: https://github.com/diux-dev/imagenet18

Ultimate Machine Learning Cheat Sheet

Notes on top-level topics from Stanford's CS 229 by Shervine Amidi and Afshine Amidi:

* Supervised learning
* Unsupervised learning
* Deep learning
* Tips and tricks
* Probability and stats refresher
* Algebra and calculus refresher

Forward this message to your Saved Messages to make sure, you won’t lose it.

Repo link: https://github.com/afshinea/stanford-cs-229-machine-learning

#Stanford #cheatsheet

Most recent version of Andrew Ng’s book Machine Learning Yearning

Link: https://gallery.mailchimp.com/dc3a7ef4d750c0abfc19202a3/files/5dd91615-3b3f-4f5d-bbfb-4ebd8608d330/Ng_MLY01_13.pdf

#andrewng #MLYearning

CACTUs: an unsupervised learning algorithm that learns to learn tasks constructed from unlabeled data. Leads to significantly more effective downstream learning & enables few-shot learning *without* labeled meta-learning datasets

ArXiv: https://arxiv.org/abs/1810.02334

#cactus #unsupervised

Hitchhiker’s guide to Exploratory Data Analysis

Exploratory Data Analysis — stage of finding out distribution of the data, volume, number of missing values and all the other characteristics of the available dataset.

Part 1: https://towardsdatascience.com/hitchhikers-guide-to-exploratory-data-analysis-6e8d896d3f7e
Part 2: https://towardsdatascience.com/hitchhikers-guide-to-exploratory-data-analysis-part-2-36ab72201e1d

#ExploratoryDA #novice #entrylevel

The Code for Facial Identity in the Primate Brain

This paper showed that facial images can be reconstructed from a simple linear model using responses of only ~200 visual neurons recorded from a monkey. This approach uses "face cells" which are encoding how much a face differs from average in particular ways ("eigenface dimensions").

https://www.sciencedirect.com/science/article/pii/S009286741730538X

#cv #dl