DMI Technical Reports: Recent submissions
Now showing items 120 of 46

Nonlinear Knowledge in Kernel Approximation
(2006)Prior knowledge over arbitrary general sets is incorporated into nonlinear kernel approximation problems in the form of linear constraints in a linear program. The key tool in this incorporation is a theorem of the ... 
Multiple Instance Classification via Successive Linear Programming
(2005)The multiple instance classification problem [6,2,12] is formulated using a linear or nonlinear kernel as the minimization of a linear function in a finite dimensional (noninteger) real space subject to linear and bilinear ... 
Feature Selection in kMedian Clustering
(2004)An e ective method for selecting features in clustering unlabeled data is proposed based on changing the objective function of the standard kmedian clustering algorithm. The change consists of perturbing the objective ... 
Breast Tumor Susceptibility to Chemotherapy via Support Vector Machines
(2003)Support vector machines (SVMs), utilizing RNA signature measurements, were used to generate a classi er to distinguish breast cancer patients that are partialresponders to chemotherapy treatment, from patients that are ... 
KnowledgeBased Linear Programming
(2003)We introduce a class of linear programs with constraints in the form of implications. Such linear programs arise in support vector machine classi cation, where in addition to explicit datasets to be classi ed, prior knowledge ... 
KnowledgeBased Nonlinear Kernel Classi ers
(2003)Prior knowledge in the form of multiple polyhedral sets, each belonging to one of two categories, is introduced into a reformulation of a nonlinear kernel support vector machine (SVM) classi er. The resulting formulation ... 
Support Vector Machine Classi cation via Parameterless Robust Linear Programming
(2003)We show that the problem of minimizing the sum of arbitrarynorm real distances to misclassi ed points, from a pair of parallel bounding planes of a classi cation problem, divided by the margin (distance) be tween the ... 
A Newton Method for Linear Programming
(2002)A fast Newton method is proposed for solving linear programs with a very large ( 106) number of constraints and a moderate ( 102) number of variables. Such linear programs occur in data mining and machine learning. ... 
Finite Newton Method for Lagrangian Support Vector Machine Classi cation
(2002)An implicit Lagrangian [19] formulation of a support vector machine classi er that led to a highly e ective iterative scheme [18] is solved here by a nite Newton method. The proposed method, which is extremely fast and ... 
An optimization approach for radiosurgery treatment planning
(20011106)We outline a new approach for radiosurgery treatment planning, based on solving a series of optimization problems. We consider a speci c treat ment planning problem for a specialized device known as the Gamma Knife, ... 
A Finite Newton Method for Classi cation Problems
(2001)A fundamental classi cation problem of data mining and machine learning is that of minimizing a strongly convex, piecewise quadratic function on the ndimensional real space Rn. We show nite termination of a Newton ... 
Set Containment Characterization
(2001)Characterization of the containment of a polyhedral set in a closed halfspace, a key factor in generating knowledgebased support vector machine classi ers [7], is extended to the following: (i) Containment of one ... 
KnowledgeBased Support Vector Machine Classi ers
(2001)Prior knowledge in the form of multiple polyhedral sets, each belonging to one of two categories, is introduced into a reformulation of a linear support vector machine classi er. The resulting formulation leads to a ... 
Incremental Support Vector Machine Classi cation
(2001)Using a recently introduced proximal support vector ma chine classi er [4], a very fast and simple incremental support vector machine (SVM) classi er is proposed which is capable of modifying an existing linear classi ... 
CrossValidation, Support Vector Machines and Slice Models
(2001)We show how to implement the crossvalidation technique used in ma chine learning as a slice model. We describe the formulation in terms of support vector machines and extend the GAMS/DEA interface to allow for e cient ... 
Data Mining via Support Vector Machines
(2001)Support vector machines (SVMs) have played a key role in broad classes of problems arising in various elds. Much more recently, SVMs have become the tool of choice for problems arising in data classi  cation and mining. ... 
SIMULATION OPTIMIZATION BASED ON A HETEROGENEOUS COMPUTING ENVIRONMENT
(2001)We solve a simulation optimization using a deterministic nonlinear solver based on the samplepath concept. The method used a quadratic model built from a collection of surrounding simulation points. The scheme does not ... 
SurvivalTime Classi cation of Breast Cancer Patients
(2001)The identi cation of breast cancer patients for whom chemother apy could prolong survival time is treated here as a data mining prob lem. This identi cation is achieved by clustering 253 breast cancer patients into ... 
Radiosurgery Treatment Planning via Nonlinear Programming
(200101)The Gamma Knife is a highly specialized treatment unit that pro vides an advanced stereotactic approach to the treatment of tumors, vascular malformations, and pain disorders within the head. Inside a shielded treatment ... 
Slice Models in General Purpose Modeling Systems
(20001214)Slice models are collections of mathematical programs with the same structure but di erent data. Examples of slice models appear in Data Envelopment Analysis, where they are used to evaluate e ciency, and crossvalidation, ...