LME Music Embedding Demo

 

Playlist Seed: please click a point
Sequencial Coherence (alpha = 1.0): large steps small steps
Popularity Bias (beta = 1.0): obscure popular
Diversity (gamma = 0.2): drift away from seed stay close to seed
Number of Songs (10): fewer more
Playlist Construction (deterministic): max probability stochastic
For details, see equation 16:
\Pr(p^{[i]}|p^{[i-1]},p^{[0]}) = \frac{e^{-\alpha \Delta(p^{[i]},~p^{[i-1]})^2 + \beta b_i - \gamma \Delta(p^{[i]},~p^{[0]})^2}}{Z(p^{[i-1]},~p^{[0]},~\alpha,~\beta,~\gamma)}
from:

Shuo Chen, Joshua Moore, Douglas Turnbull, Thorsten Joachims
Playlist Prediction via Metric Embedding
ACM Conference on Knowledge Discovery and Data Mining (KDD)
August 2012