Day 35: Human-AI Collaborations

Linear regressions with 2nd/3rd order polynomials were equal to or better than neural-nets (aka 'deep learning') in every data set they looked at. https://t.co/e0DTbtaKDa

— David Baranger (@DABaranger) June 23, 2018

Automatically removing raindrops from an image. Useful for self-driving cars. Trains an attentive generative network using adversarial training. Learns about raindrop regions & their surroundings, then generates focused on surroundings. https://t.co/y423rQkMdk pic.twitter.com/e6FgxXyBSl

— Reza Zadeh (@Reza_Zadeh) June 23, 2018

Our survey paper on Procedural Content Generation via Machine Learning (PCGML) is now officially published. We survey the nascent field of using ML to generate game content.

Available as early access on IEEE Xplore:https://t.co/AhfLbbUI4J
And on ArXiv:https://t.co/riV6eYuCZ2 pic.twitter.com/xVno9TmS0b

— Julian Togelius (@togelius) June 13, 2018

GrCAN: Gradient Boost Convolutional Autoencoder with Neural Decision Forest. https://t.co/2eqxqwDfDX pic.twitter.com/XN7hWfLjPa

— arxiv (@arxiv_org) June 23, 2018

Human-AI Collaborations in 1,500 companies: training, explaining, interacting, and more: https://t.co/1JLaaOqoSE by @pauldaugh

— Oren Etzioni (@etzioni) June 23, 2018

"Data science vs. statistics: two cultures?" A good review article collecting & synthesizing all the diff opinions over the years. If someone should still ask what data science is about, that's the article to refer them to https://t.co/TPy5UEHTfI

— Sebastian Raschka (@rasbt) June 23, 2018

Also has a good set of references on when/how the term was coined: pic.twitter.com/fqssPaOb28

— Sebastian Raschka (@rasbt) June 23, 2018

This post about missing values in #julia gives a good overview of the “software engineer’s null” vs. the “data scientist’s null” and why it matters. ht @clarkfitzg https://t.co/AeDqym7Z6R

— Jenny Bryan (@JennyBryan) June 23, 2018

My Do More With R series of short screencasts so far:

* Interactive scatter plots
* dplyr's case_when
* test code with testthat
* easy #rstats dashboards
* @RStudio code snippetshttps://t.co/Ex2LlfjNwF#R4DS pic.twitter.com/wzNxW6m7X4

— Sharon Machlis (@sharon000) June 23, 2018

A nice review article on recent advances in variational inference (C. Zhang, J. Butepage, H. Kjellstrom, S. Mandt): https://t.co/SLVVPAmchE #MachineLearning

— Diana Cai (@dianarycai) June 23, 2018

New post:
Small n correlations + p values = disasterhttps://t.co/KBgRHGCvLg pic.twitter.com/mBe2oRb3Tk

— Guillaume Rousselet (@robustgar) June 22, 2018

Dissolving the Fermi Paradox: why taking uncertainties into account makes the absence of aliens less weird, and why an empty sky doesn't foretell our doom. https://t.co/7C6g14Spbd Popular FAQ: https://t.co/3a6piKhAmi

— Anders Sandberg (@anderssandberg) June 23, 2018

Deep learning is useful because it enables us to create programs that we could not otherwise code by hand. But the space of programs you can learn via deep learning models is a minuscule slice of the space of programs that we may be interested in.

— François Chollet (@fchollet) June 22, 2018