INSIGHTS:
How does semantic distance affect the shortest paths found by A*, when compared to other heuristics for the shortest paths? -> (maybe try with looking first for category related / nº of links pointing at the successor)
If time allows, are there any superior AI methods for exploring the graph?
Can we identify any patterns or recurring structures in the human paths (ie. going for a central hub)?
What insights can we draw from the cases where human paths outperform the machines ones? And viceversa?
Are there any specific Wikipedia categories that are correlated with the differences in the performances?
How do these compare to the optimal path?
How many nodes does the machine go through and how many does the human go through?
https://eudl.eu/pdf/10.4108/icst.collaboratecom.2011.247162
Idea of using landmarks could lead to a very scalable solution, and we have arguments behind this being valid. It's also an alternate approach that focus on getting graph knowledge instead of just embeddings
INSIGHTS: How does semantic distance affect the shortest paths found by A*, when compared to other heuristics for the shortest paths? -> (maybe try with looking first for category related / nº of links pointing at the successor) If time allows, are there any superior AI methods for exploring the graph? Can we identify any patterns or recurring structures in the human paths (ie. going for a central hub)? What insights can we draw from the cases where human paths outperform the machines ones? And viceversa? Are there any specific Wikipedia categories that are correlated with the differences in the performances? How do these compare to the optimal path? How many nodes does the machine go through and how many does the human go through?
Things to do:
Thursday 14 meeting (?) algos
LINKS FROM THE MEETING: https://github.com/mohakbhardwaj/SaIL/tree/ed88809893022342e9354a819ba8a9dab2bfeb5d https://github.com/mohakbhardwaj/SaIL/blob/ed88809893022342e9354a819ba8a9dab2bfeb5d/SaIL/agents/sail_agent.py#L23 single_source_shortest_path: https://networkx.org/documentation/stable/reference/algorithms/generated/networkx.algorithms.shortest_paths.unweighted.single_source_shortest_path.html#networkx.algorithms.shortest_paths.unweighted.single_source_shortest_path Source code for networkx.algorithms.shortest_paths.unweighted: https://networkx.org/documentation/stable/_modules/networkx/algorithms/shortest_paths/unweighted.html#single_source_shortest_path Finding shortest paths with Graph Neural Networks: https://medium.com/octavian-ai/finding-shortest-paths-with-graph-networks-807c5bbfc9c8 t-distributed stochastic neighbor embedding: https://en.wikipedia.org/wiki/T-distributed_stochastic_neighbor_embedding shortest path another way: https://medium.com/octavian-ai/finding-shortest-paths-with-graph-networks-807c5bbfc9c8