How AI shaped online shopping in the 2010s

ht Guilherme Carrara Neto

Quote of the day - machine learning

Assorted Links

1. Paper: "exposure to socioeconomic inequality in an everyday setting negatively affects willingness to publicly support a redistributive economic policy" by Melissa Sands
   
   
2. Peak Artificial Intelligence
   
   
3. Interactive historic map of every UK railway line that's ever existed, via Tim Dunn
   
   
4. Oxford DEMIG databases with unprecedented coverage of migration & migration policy data now available for free download, via Hein de Haas
   
   
5. Using machine learning in R to predict Automobile Collision and Bike Share Demand and subsidy via Ken Steif
   
   
6. Awful AI: a curated list to track current scary usages of AI - hoping to raise awareness, HT Will Geary
   
   
7. The software used to create the black hole in the movie 'Interstellar' is a full implementation of Einstein's equations in 40,000 lines of C++, and rendered thousands of 23-megapixel IMAX frames on a 32,000-core render farm at about 20 core-hours per frame via John Arundel
   
   
8. What to consider when choosing colors for data visualization, by Lisa Rost

credit: Lisa Rost

Using deep learning and satellite imagery to improve land use classification in cities

Marta Gonzalez and colleagues have a recent paper using deep learning and satellite image data to improve land use classification. The authors have made documented code and Jupyter notebooks available here. I'm self recommitting the paper and code to my future self. HT Marco De Nadai.

Albert, A., Kaur, J., & Gonzalez, M. C. (2017, August). Using convolutional networks and satellite imagery to identify patterns in urban environments at a large scale. In Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 1357-1366). ACM.

Abstact:

Urban planning applications (energy audits, investment, etc.) require an understanding of built infrastructure and its environment, i.e., both low-level, physical features (amount of vegetation, building area and geometry etc.), as well as higher-level concepts such as land use classes (which encode expert understanding of socio-economic end uses). This kind of data is expensive and labor-intensive to obtain, which limits its availability (particularly in developing countries). We analyze patterns in land use in urban neighborhoods using large-scale satellite imagery data (which is available worldwide from third-party providers) and state-of-the-art computer vision techniques based on deep convolutional neural networks. For supervision, given the limited availability of standard benchmarks for remote-sensing data, we obtain ground truth land use class labels carefully sampled from open-source surveys, in particular the Urban Atlas land classification dataset of $20$ land use classes across $~300$ European cities. We use this data to train and compare deep architectures which have recently shown good performance on standard computer vision tasks (image classification and segmentation), including on geospatial data. Furthermore, we show that the deep representations extracted from satellite imagery of urban environments can be used to compare neighborhoods across several cities. We make our dataset available for other machine learning researchers to use for remote-sensing applications.

R Links

1. Text Mining with R
   
   
2. A great Shiny app to deal with the Traveling Salesman problem
   
   
   
   
   
3. Nice color palettes based on Wes Anderson movies, by Karthik Ram
   
   
4. Rstudio Gallery of examples using TensorFlow and R for deep learning
   
   
5. Cohort analysis: Retention Rate Visualization with R
   
   
6. Network analysis of Game of Thrones by Shirin Glander, via Mara Averick
   
   
7. Creating interactive plots in R with plotly
   
   
8. How to create a smooth transition between chloropleth and cartogram in R

credit: Yan Holts

Using deep learning and Google Street View to estimate the socioeconomic characteristics of neighborhoods

Timnit Gebru and her colleagues have recently published a very interesting paper in PNAS. The authors developed a method to analyze car images from Google Street View using machine-learning and computer vision methods to determine socioeconomic statistics and political preferences at the zip code level in the US. There is a video of Timnit Gebru presenting the paper here. 

Gebru, T., Krause, J., Wang, Y., Chen, D., Deng, J., Aiden, E. L., & Fei-Fei, L. (2017). Using deep learning and Google Street View to estimate the demographic makeup of neighborhoods across the United States. PNAS 2017 : 1700035114v1-201700035 [arxiv version here]

Abstract:

The United States spends more than $250 million each year on the American Community Survey (ACS), a labor-intensive door-to-door study that measures statistics relating to race, gender, education, occupation, unemployment, and other demographic factors. Although a comprehensive source of data, the lag between demographic changes and their appearance in the ACS can exceed several years. As digital imagery becomes ubiquitous and machine vision techniques improve, automated data analysis may become an increasingly practical supplement to the ACS. Here, we present a method that estimates socioeconomic characteristics of regions spanning 200 US cities by using 50 million images of street scenes gathered with Google Street View cars. Using deep learning-based computer vision techniques, we determined the make, model, and year of all motor vehicles encountered in particular neighborhoods. Data from this census of motor vehicles, which enumerated 22 million automobiles in total (8% of all automobiles in the United States), were used to accurately estimate income, race, education, and voting patterns at the zip code and precinct level. (The average US precinct contains ∼∼1,000 people.) The resulting associations are surprisingly simple and powerful. For instance, if the number of sedans encountered during a drive through a city is higher than the number of pickup trucks, the city is likely to vote for a Democrat during the next presidential election (88% chance); otherwise, it is likely to vote Republican (82%). Our results suggest that automated systems for monitoring demographics may effectively complement labor-intensive approaches, with the potential to measure demographics with fine spatial resolution, in close to real time.

cities used to train a model estimating socioeconomic data from car attributes 

Related posts:

• Using Computer Vision to Measure the Quality and Impact of Urban Appearance

• The Impact of Urban Appearance

Assorted Links

1. Isoscope: work by Flavio Gortana (Twitter), showing how isochrones by car vary across space and time
   
2. Tutorial: Spatial Datasets and Urban Applications (code, slides + data)
   
   
3. Beautiful photographs of Forgotten Places, via Darran Anderson
   
   
4. Five tools that help you master map projections
   
   
5. The unfinished cities of Spain via Urban Living Lab
   
   
6. 15 years of urban growth in China, via Lazaro Gamio
   
   
7. Surnames and ancestry in Brazil, by Leo Monasterio
   
   
8. "Seeing Theory – A visual introduction to probability and statistics.”
   
   
9. paper: Combining satellite imagery and machine learning to predict poverty, by Neal Jean et al [code]

The Impact of Urban Appearance

We have mentioned the work of Cesar Hidalgo (twitter) in this blog a few times already but he keeps coming up with new interesting research . As many will know, Hidalgo and his team have a project called Place Pulse, in which they use computer vision techniques to analyze snapshots of google street view and crowdsource people's perception on city’s physical appearance. Here are my three highlight of the project so far.

• In its initial phase, the project focused on 19 American cities. Some of the research findings show how perceived safety of streetscapes has strong positive correlation with household income, population density and income inequality (PDF of the paper).
  
  
• In a more recent paper, led by Marco De Nadai (twitter), the authors combine Place Pulse with mobile phone data to show how the safety scores of neighbourhoods in Rome and Milan are also strongly correlated with urban liveliness, and that visual features like greenery and street facing windows contribute to a positive appearance of safety. This is an extension of Nadai's previous work, where he engages more directly with Jane Jacobs's ideas on urban vitality.
  
  
• Finally, the Place Pulse project has been recently updated and expanded to create the first global dataset of urban appearance "...containing 110,988 images from 56 cities, and 1,170,000 pairwise comparisons provided by 81,630 online volunteers along six perceptual attributes: safe, lively, boring, wealthy, depressing, and beautiful". 

The research possibilities here are endless, so I'm sure I'll be posting updates on this project again in the future. Don't forget to input some of your own perceptions to the project.

Using Computer Vision to Measure the Quality and Impact of Urban Appearance

Interesting paper on computer vision & urban economics, by Cesar Hidalgo and colleagues. Ungated version here.

Naik, N., Raskar, R., & Hidalgo, C. A. (2016). Cities Are Physical Too: Using Computer Vision to Measure the Quality and Impact of Urban Appearance. The American Economic Review, 106(5), 128-132.

Abstract

For social scientists, developing an empirical connection between the physical appearance of a city and the behavior and health of its inhabitants has proved challenging due to a lack of data on urban appearance. Can we use computers to quantify urban appearance from street-level imagery? We describe Streetscore: a computer vision algorithm that measures the perceived safety of streetscapes. Using Streetscore to evaluate 19 American cities, we find that the average perceived safety has a strong positive correlation with population density and household income; and the variation in perceived safety has a strong positive correlation with income inequality.

Assorted Links

1. [new paper] Mortality differentials by education in Brazil: Life expectancy was 6.27 years longer for men with a complete university degree than for those with less than primary education in 2010. The paper is in Portuguese
   
   
2. A nice summary of 10 major data science methods
   
   
3. Beautiful photos of 1908 Japan
   
   
4. 40% of the buildings in Manhattan could not be built today under current zoning rules
   
   
5. Will autonomous vehicles be safer then conventional cars ? Here is why answering this question is far from simple  via Tyler Cowen
   
   
6. Lisa Charlotte Rost ‏ has gone through the painful task of creating the same chart with 12 different tools and libraries. This gives me more reasons to stick to R and ggplot2 :)
   
   
7. Road intersections as art
   
   
8. Making explicit the network behind air transport: by Martin Grandjean

Assorted Links

1. Taking Data Visualization From Eye Candy to Efficiency
   
   
2. Fuck Nuance, a manifesto for more bold social theories, by Kieran Healy (via Analise Real)
   
   
3. Machine Learning Classes at Oxford (videos, presentation and code)
   
   
4. 2015 Ig Nobel prizes: a study shows that nearly all mammals take the same amount of time to urinate (about 21 seconds)
   
   
5. There is a Space Syntax Toolkit for QGIS (ht Renato Saboya, who runs a great blog about urbanism written in Portuguese: Urbanidades)
   
   
6. 10 things you need to know about the Magna Carta
   
   
7. Slum tourism and its discontents. Not exactly a new article but a good summary of arguments (ht Fabiana Freitas, my lovely partner)
   
   
8. The Value of Democracy: Evidence from Road Building in Kenya [unagated paper]
   
   
9. The Human Mortality Explorer. Visualizing Death across Time and Space via (Bernardo Queiroz)