Codes

Most of the codes related to my papers can be accessed below. Contact me if you don’t find something.

• SPOTgreedy for prototype selection with optimal transport theory. The Python version is available at here.
• ROT4C code for classification with robust optimal transport loss.
• Generating geometry-aware word meta embeddings.
• MBA code for unsupervised cross-lingual word alignment.
• RASA code for Riemannian adaptive stochastic algorithm. The python version is available at here.
• GitHub repo on Riemannian stochastic optimization solvers.
• McTorch, a library for manifold constrained deep learning.
• GeoMM to learn multilingual word embeddings.
• Domain adaptation using metric learning on manifolds.
• Side information infused incremental tensor analysis.
• Low-rank geometric mean metric learning.
• A dual framework for tensor completion.
• Structured matrix learning.
• R-SVRG, accelerated stochastic gradients on manifolds.
• Decentralized subspace learning with gossiping.
• Sparse and low-rank index coding with Riemannian optimization.
• Scaled SGD for low-rank matrix completion.
• Symmetry-invariant optimization in deep networks.
• Riemannian preconditioning for tensor completion.
• Sparse plus low-rank autoregressive identification in neuroimaging time series.
• A Riemannian approach to low-rank algebraic Riccati equations.
• R3MC: A Riemannian three-factor algorithm for low-rank matrix completion.
• qGeomMC: A quotient geometric approach to low-rank matrix Completion.
  • It also contains a number of computationally efficient Riemannian algorithms for fixed-rank (low-rank) matrix completion for different matrix factorizations.
• Trace norm minimization of the form f(X) + λ || X ||_*, where f is a smooth convex function, || X ||_* is the trace (or nuclear) norm, and λ >= 0 is a regularization parameter.
• Euclidean distance matrix (EDM) completion.