Day 53: Talk the Walk

The tortuous paths that France and Croatia took to the final https://t.co/k5q05jzJmF pic.twitter.com/LGoo0DyAVF

— Ritchie King (@RitchieSKing) July 11, 2018

This is a neat way to visualize a college curriculum. #education #dataviz

Bonus: Could change the color/size of the circles to indicate grades achieved for that course.https://t.co/u6oJEJ1kp5 pic.twitter.com/iXM1HiYwPh

— Randy Olson (@randal_olson) July 11, 2018

I recreated an NYT graphic (made in D3) in 10 minutes using ggparliament, a #ggplot2 #tidyverse extension in #rstats. H/T to @thosjleeper, @robert_squared, and uhhh... myself for doing a lot of work on ggparliament! pic.twitter.com/Tzs37DJ8zY

— Zoe Meers (@zoe_meers) July 12, 2018

I did a comparison of city street network orientations in major US cities, and now I've got a better sense of why I find Boston so difficult to navigate. Visualization uses Python, OSMnx, and @OpenStreetMap data.https://t.co/prINZbDh9z pic.twitter.com/YGY4VDSjZY

— Geoff Boeing (@gboeing) July 11, 2018

All our #datajournalism is open source and we publish the raw data on @github https://t.co/j8BLcUrxAF#OpenData #DDJ pic.twitter.com/f7mh3esk2M

— Simon Rogers (@smfrogers) July 11, 2018

Learning General Purpose Distributed Sentence Representations via Large Scale Multi-task Learning #ICLR2018

General purpose, fixed-length representations of sentences via multi-task training by @sandeep1337 @APTrizzle

ArXivhttps://t.co/WkUoadjRDn
Githubhttps://t.co/fCPuFBmUjI pic.twitter.com/7le5BHPTGX

— ML Review (@ml_review) July 12, 2018

Our London & NYC-based #AI researchers have developed a new approach to find more relevant, related facts in #KnowledgeGraphs; learn more about their efforts and #SIGIR2018 paper https://t.co/xZBnFf1cuD pic.twitter.com/McprlLVRk9

— Tech At Bloomberg (@TechAtBloomberg) July 11, 2018

"Talk the Walk" a dataset and environment for AI agent to teach each other visual navigation.
Brought to you by FAIR. https://t.co/3CcaAOXZYH

— Yann LeCun (@ylecun) July 11, 2018

Measuring abstract reasoning in neural networks
"the model's ability to generalise improves markedly if it is trained to predict symbolic explanations for its answers"https://t.co/wWhxlWQmR7#DeepMind pic.twitter.com/oa8bRK4iYZ

— Thomas (@evolvingstuff) July 12, 2018

Interesting that DeepMind is working on a IQ-like test to measure abstract reasoning capabilities. I've been working on a very similar benchmark for the past 6 months (taking a more formal approach). Good to see more action in that space (cc @ChrSzegedy) https://t.co/zGXrdekF70

— François Chollet (@fchollet) July 11, 2018

(arxiv) Experiments with Variational Attention (https://t.co/IB51FIdjTt, https://t.co/gwt7uZszAg): fast training for latent var. attention (e.g. hard attention) with accuracy like soft attention. Learns sharp attention distributions (red), and useful posterior corrections (blue) pic.twitter.com/M8rS3B3fqK

— harvardnlp (@harvardnlp) July 11, 2018

DALEX: Descriptive Machine Learning Explanations for Black Box Models #rstats https://t.co/KrSElIvycU pic.twitter.com/KtVcCZYddg

— ML Review (@ml_review) July 11, 2018

Released lolviz 1.4 that can visualize 1D, 2D numpy arrays in jupyter or raw Python 2, 3. Use "pip install -U lolviz" to enjoy. https://t.co/L3HLISC4lS pic.twitter.com/3VJVwaFwBK

— Terence Parr (@the_antlr_guy) July 11, 2018

Another great resource from @AmeliaMN: explaining and visualizing / contextualizing histograms https://t.co/NkTZZBdYxE #icots10 #rstats

— Hilary Parker (@hspter) July 12, 2018

Thank you to everyone who listened in on my #SciPy2018 talk! Sorry I didn't make it through the "Future of AutoML" portion - that live demo took longer than expected. I've posted my slides online here: https://t.co/SW5Q5BS9Fn

— Randy Olson (@randal_olson) July 11, 2018

Hadley’s slides, Should all statistics students be programmers? are available here https://t.co/78N3W28Eo6 #icots10 @hadleywickham

— Amelia McNamara (@AmeliaMN) July 12, 2018

A software is used for detecting potential welfare fraudsters in the Netherlands. The government keeps quiet about how that works. Civil rights activists are taking the matter to court. Read "High-Risk Citizens" by @iljabraun (also available in German) https://t.co/lSbt4hcCUb

— AlgorithmWatch (@algorithmwatch) July 4, 2018

One of the more striking examples I've seen of an algorithm solving the wrong problem pic.twitter.com/tMVpC54RlJ

— Janelle Shane (@JanelleCShane) July 9, 2018

I'd go further - any feedback system that does not itself accept feedback will not ultimately help teachers get better at teaching. https://t.co/A3tGj4APfj

— Cathy O'Neil (@mathbabedotorg) July 12, 2018

Exactly, these are the issues! Quadrigram still isn’t really right. All covered in the original quote—except the omission of five-grams does make the piece show its age. pic.twitter.com/mjs8S9aSSm

— Christopher Manning (@chrmanning) July 12, 2018