Signed Spectral Clustering

Implementation of the Signed Normalized Cuts algorithm described in Gallier (2016).

Results

Figure 1: Sample ingredient clusters

Figure 2: Ingredient affinity matrix computed by pairwise normalized PMI (NPMI) scores

Organization

• ssc.py contains the signed normalized cuts implementation
• affinities.py generates an affinity matrix from a recipe dataset, clusters the ingredients, and performs various recipe rating prediction experiments using the clusters as features.

See Report.pdf for a thorough discussion of the implementation and application to recipe rating predictions.