Dynamic Point Selection for the L1 Norm Hyperplane Separation Problem
نویسندگان
چکیده
L1 norm discrimination consists in finding the hyperplane that minimizes the sum of L1 norm distances between the hyperplane and the points that lie on the wrong side of the hyperplane. This problem is difficult for datasets containing more than 100,000 points. Since few points are needed to obtain the optimal hyperplane, we propose a point selection algorithm which iteratively adds the necessary points to a reduce set of points. We evaluate different point selection criteria. The resulting method obtains in reasonable times exact solutions for problems of up to two million points.
منابع مشابه
On the Problem of Finding the Least Number of Features by L1-Norm Minimisation
Recently, the so-called Support Feature Machine (SFM) was proposed as a novel approach to feature selection for classification. It relies on approximating the zero-norm minimising weight vector of a separating hyperplane by optimising for its one-norm. In contrast to the L1-SVM it uses an additional constraint based on the average of data points. In experiments on artificial datasets we observe...
متن کاملArbitrary-norm Separation by Variable Neighborhood Search
We consider the problem of separating two sets of points in an Euclidean space with a hyperplane that minimizes the sum of p-norm distances to the plane of points lying on the “wrong” side of the plane. A Variable Neighborhood Search metaheuristic framework is used to determine the plane coefficients. For a set of examples with L1-norm, L2-norm and L∞-norm, for which the exact solution can be c...
متن کاملTightened L 0 - Relaxation Penalties for Classification 1 by Noam Goldberg , 2 Jonathan Eckstein
In optimization-based classification model selection, for example when using linear programming formulations, a standard approach is to penalize the L1 norm of some linear functional in order to select sparse models. Instead, we propose a novel integer linear program for sparse classifier selection, generalizing the minimum disagreement hyperplane problem whose complexity has been investigated ...
متن کاملAn L1-norm method for generating all of efficient solutions of multi-objective integer linear programming problem
This paper extends the proposed method by Jahanshahloo et al. (2004) (a method for generating all the efficient solutions of a 0–1 multi-objective linear programming problem, Asia-Pacific Journal of Operational Research). This paper considers the recession direction for a multi-objective integer linear programming (MOILP) problem and presents necessary and sufficient conditions to have unbounde...
متن کاملApproximate L0 constrained non-negative matrix and tensor factorization
Non-negative matrix factorization (NMF), i.e. V ≈ WH where both V, W and H are non-negative has become a widely used blind source separation technique due to its part based representation. The NMF decomposition is not in general unique and a part based representation not guaranteed. However, imposing sparseness both improves the uniqueness of the decomposition and favors part based representati...
متن کامل