We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Practical Algorithms for Learning Near-Isometric Linear Embeddings

Abstract: We propose two practical non-convex approaches for learning near-isometric, linear embeddings of finite sets of data points. Given a set of training points $\mathcal{X}$, we consider the secant set $S(\mathcal{X})$ that consists of all pairwise difference vectors of $\mathcal{X}$, normalized to lie on the unit sphere. The problem can be formulated as finding a symmetric and positive semi-definite matrix $\boldsymbol{\Psi}$ that preserves the norms of all the vectors in $S(\mathcal{X})$ up to a distortion parameter $\delta$. Motivated by non-negative matrix factorization, we reformulate our problem into a Frobenius norm minimization problem, which is solved by the Alternating Direction Method of Multipliers (ADMM) and develop an algorithm, FroMax. Another method solves for a projection matrix $\boldsymbol{\Psi}$ by minimizing the restricted isometry property (RIP) directly over the set of symmetric, postive semi-definite matrices. Applying ADMM and a Moreau decomposition on a proximal mapping, we develop another algorithm, NILE-Pro, for dimensionality reduction. FroMax is shown to converge faster for smaller $\delta$ while NILE-Pro converges faster for larger $\delta$. Both non-convex approaches are then empirically demonstrated to be more computationally efficient than prior convex approaches for a number of applications in machine learning and signal processing.
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG); Optimization and Control (math.OC)
MSC classes: 90C90
Cite as: arXiv:1601.00062 [stat.ML]
  (or arXiv:1601.00062v2 [stat.ML] for this version)

Submission history

From: Hao-Jun Shi [view email]
[v1] Fri, 1 Jan 2016 09:06:11 GMT (395kb,D)
[v2] Fri, 22 Apr 2016 21:47:56 GMT (488kb,D)

Link back to: arXiv, form interface, contact.