Neural Embeddings for Populated Geonames Locations thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Neural Embeddings for Populated Geonames Locations

Published on Nov 28, 2017746 Views

The application of neural embedding algorithms (based on architectures like skip-grams) to large knowledge bases like Wikipedia and the Google News Corpus has tremendously benefited multiple com- muni

Related categories

Chapter list

Neural Embeddings for Populated GeoNames Locations00:00
Motivation: feature extraction from locations00:08
Machine learning applications01:42
Motivation: feature extraction from locations03:53
What makes for a ‘good’ feature space?04:30
Do lat-long points capture proximity semantics?07:04
More formally...07:46
Do lat-long points capture proximity semantics?08:49
Do lat-long points capture proximity semantics? - 109:26
What makes for a good feature space?10:25
Idea: ‘Embed’ Geonames as a weighted, directed network...11:03
Step 1: Determine set of nodes in network13:35
Step 2: Determine edges and weights14:21
Step 3: Run DeepWalk on network15:25
Example in paper: North Dakota15:54
Vectors, code and raw data all on GitHub (also, figshare)16:25