Artificiality Bites 💊 Issue #17

Hello Human! This is a new issue from my weekly newsletter, holding a tiny compilation made of interesting articles from last week, projects, tutorials and tools; all related to Data, Artificial Intelligence and adjacent topics. Bon appetit!

📝 Interesting articles this week

• The Machine Learning Behind Hum to Search 10'
  Launched in October, Hum to Search is a new fully machine-learned system within Google Search that allows a person to find a song using only a hummed rendition of it. This article explains the magic behind that.
• Alexa gets better at predicting customers’ goals 6'
  With a new machine learning system, Alexa can now infer that an initial question implies a subsequent request.
• Turning Metadata Into Insights with Databook 26'
  Databook is Uber’s internal platform built to surface and manage the metadata related to various data entities. This article describes how Uber evolved this platform in order to make it scale as their systems became larger and more complex.

💡 Projects / Papers

• Graph Kernels — State-of-the-Art and Future Challenges 📝 151p
  This publication provides a review of existing graph kernels, their applications, software plus data resources, and an empirical comparison of state-of-the-art graph kernels.

🔧 Tutorials

• Leveraging Pre-trained Language Model Checkpoints for Encoder-Decoder Models 71'
  How encoder-decoder models can be warm-started, to skip the costly pre-training with competitive results. Contains a complete code example using Transformers.
• Hierarchical Time Series with Prophet and PyMC3 📹 63'
  How to build a hierarchical version of Facebook’s Prophet package to make long-term predictions for multiple, related, time-series.
• OpenCV Super Resolution with Deep Learning 27'
  How to achieve super resolution in images and real-time video streams using OpenCV and Deep Learning.
• I Built an AI Stylist Inspired by Social Media 13'
  How to choose an outfit with AI, based on fashion influencers' looks on Instagram and combining them with pictures of clothing you already own.

📦 Repositories

• tzt101/MichiGAN
  Multi-Input-Conditioned Hair Image Generation for Portrait Editing.
• facebookresearch/pifuhd
  This repository contains a Pytorch implementation of “Multi-Level Pixel-Aligned Implicit Function for High-Resolution 3D Human Digitization”, starting from one single image.
• google-research-datasets/Objectron/
  Objectron is a dataset of short object-centric video clips capturing a huge set of common objects from different angles, with pose annotations. Read more at Google AI Blog.

🎓 Courses / Events

• The Private AI Series
  A new course starting in January where you’ll learn how privacy is impacting every domain and how to build real-world products with privacy-preserving AI technology. Offered for free by PyTorch, the University of Oxford and OpenMined.
• PyTorch Developer Day 2020 5h18'
  The PyTorch Developer Day took place on November 12, including technical talks about updates and tools to support development across a variety of domains, and covering the latest research around systems and tooling in ML.

🚀 Extra bits

• How AI predictions fared against pollsters in the 2020 U.S. election 6'
• A.I. attempts to play Liszt’s Hungarian Rhapsody 1'

(How to deceive a machine learning algorithm)

👉 Newsletter en Español
👋 See you next week!