Search

Ioannis Akrotirianakis

from Princeton, NJ
Age ~56
Get Report
Also known as:
Loannis Akrotirianaki
Phone and address:
37B Melrose Ct, Princeton, NJ 08540
Connected to:
Ahmad A Akrouche, 36Allen M Akroush, 48Osama A Akroush, 42Sami J Akroush, 80Zoubir Akrout, 77Margo J Akroyd, 74Anastasia O Akruzhnou, 38Leanne Aks, 30Sharon L Aks, 66Cody Aksamit, 29

Ioannis Akrotirianakis Phones & Addresses

  • 37B Melrose Ct, Princeton, NJ 08540
  • Fort Dix, NJ
  • Plainsboro, NJ
  • Cary, NC

Work

Company: Siemens Jan 2015 Position: Senior scientist at siemens

Education

Degree: Doctorates, Doctor of Philosophy School / High School: Dyson School of Design Engineering 1995 to 2000 Specialities: Computer Science

Skills

Algorithms • Mathematical Modeling • Operations Research • Data Mining • Machine Learning • Statistics • Data Analysis • Optimization • C++ • Statistical Modeling • High Performance Computing • Sas • Artificial Intelligence • R • Latex

Industries

Computer Software

Resumes

Resumes

Ioannis Akrotirianakis Photo 1

Senior Scientist At Siemens

View page
Location:
37b Melrose Ct, Princeton, NJ 08540
Industry:
Computer Software
Work:
Siemens
Senior Scientist at Siemens
Siemens Oct 2010 - Dec 2014
Research Scientist
Sas Oct 2004 - Sep 2010
Operations Research Specialist
Princeton University Sep 2001 - Sep 2004
Postdoctoral Research Associate
Imperial College London Sep 1999 - Aug 2001
Research Assistant
Education:
Dyson School of Design Engineering 1995 - 2000
Doctorates, Doctor of Philosophy, Computer Science The University of Manchester 1991 - 1992
Master of Science, Masters, Computer Science Aristotle University of Thessaloniki (Auth) 1986 - 1990
Bachelors, Bachelor of Science, Mathematics
Skills:
Algorithms
Mathematical Modeling
Operations Research
Data Mining
Machine Learning
Statistics
Data Analysis
Optimization
C++
Statistical Modeling
High Performance Computing
Sas
Artificial Intelligence
R
Latex

Publications

Us Patents

Primal-Dual Interior Point Methods For Solving Discrete Optimal Power Flow Problems Implementing A Chain Rule Technique For Improved Efficiency

View page
US Patent:
20120150504, Jun 14, 2012
Filed:
Dec 5, 2011
Appl. No.:
13/310829
Inventors:
Ioannis Akrotirianakis - Plainsboro NJ, US
Andrey Torzhkov - Lafayette CA, US
Assignee:
Siemens Corporation - Iseline NJ
International Classification:
G06F 17/11
US Classification:
703 2
Abstract:
A solution to the optimal power flow (OPF) problem for electrical generation and distribution systems utilizes a re-configuration of the OPF problem that allows for a simplified analysis and resolution of a network-based OPF problem in a minimal number of iterations. The standard mixed integer quadratic problem (MIQP) definition is be reconfigured, using the chain rule, to a relatively compact linear system of six equations with six unknowns (the smallest reducible (atomic) problem). Advantageously, the reduction in the complexity of the problem does not require any assumptions and yields a solution equivalent to the original problem.

Critical Threshold Parameters For Defining Bursts In Event Logs

View page
US Patent:
20120246109, Sep 27, 2012
Filed:
Nov 1, 2011
Appl. No.:
13/286473
Inventors:
Fabian Moerchen - Rocky Hill NJ, US
Ioannis Akrotirianakis - Plainsboro NJ, US
Assignee:
Siemens Corporation - Iselin NJ
International Classification:
G06N 5/02
US Classification:
706 58
Abstract:
Systems and methods for determining critical thresholds on a number of events (k) and a window length (t) for properly defining a burst of events in a data stream. A new coverage metric Cis defined and used in the determination, where the coverage metric Cis defined for a particular pair (k,t) as a fraction, with the numerator defined a number of events that occur within some (k,t)-bursty window and the denominator defined as the total number of events (n) that occurred along the entire time span being analyzed. Coverage metric Cis monotonic non-increasing in k and monotonic non-decreasing in t, allowing for a divide-and-conquer search strategy to be used to find the critical threshold pairs (k*, t*).

Hybrid Interior-Point Alternating Directions Algorithm For Support Vector Machines And Feature Selection

View page
US Patent:
20130073489, Mar 21, 2013
Filed:
Sep 12, 2012
Appl. No.:
13/611528
Inventors:
Zhiwei Qin - New York NY, US
Xiaocheng Tang - Plainsboro NJ, US
Ioannis Akrotirianakis - Plainsboro NJ, US
Amit Chakraborty - East Windsor NJ, US
Assignee:
Siemens Corporation - Iselin NJ
International Classification:
G06F 15/18
US Classification:
706 12
Abstract:
A method for training a classifier for selecting features in sparse data sets with high feature dimensionality includes providing a set of data items x and labels y, minimizing a functional of the data items x and associated labels yto solve for hyperplane w and offset b of a classifier by successively iteratively approximating w and b, auxiliary variables a and c, and multiplier vectors γand γ, wherein λ, λ, μ, and μare predetermined constants, e is a unit vector, and X and Y are respective matrix representations of the data items x and labels y; providing non-zero elements of the hyperplane vector w and corresponding components of X and Y as arguments to an interior point method solver to solve for hyperplane vector w and offset b, wherein w and b define a classifier than can associate each data item x with the correct label y.

Interior Point Method For Reformulated Optimal Power Flow Model

View page
US Patent:
20130238148, Sep 12, 2013
Filed:
Mar 6, 2012
Appl. No.:
13/413086
Inventors:
Ruken Duzgun - Bethlehem PA, US
Ioannis Akrotirianakis - Plainsboro NJ, US
Amit Chakraborty - East Windsor NJ, US
Assignee:
Siemens Corporation - Iselin NJ
International Classification:
G05F 5/00
US Classification:
700286
Abstract:
A method for approximating an optimal power flow of a smart electric power grid includes providing a cost function that models a smart electric power grid having buses connected by branches, deriving a set of linear equations that minimize the cost function subject to constraints from an expression of an extremum of the cost function with respect to all arguments, reducing a dimension of the linear equations by solving for a subset of the linear equations, re-organizing the reduced dimension linear equations into primal and dual parts, and decomposing the re-organized reduced dimensional linear equations into two systems of block matrix equations which can be solved by a series of back substitutions. A solution of the two systems of block matrix equations yields conditions for a lowest cost per kilowatthour delivered through the smart electric power grid.

Second-Order Optimization Methods For Avoiding Saddle Points During The Training Of Deep Neural Networks

View page
US Patent:
20210357740, Nov 18, 2021
Filed:
Apr 12, 2018
Appl. No.:
16/337154
Inventors:
- Munich, DE
Ioannis Akrotirianakis - Princeton NJ, US
Amit Chakraborty - East Windsor NJ, US
International Classification:
G06N 3/08
G06K 9/62
Abstract:
A computer-implemented method for training a deep neural network includes defining a loss function corresponding to the deep neural network, receiving a training dataset comprising training samples, and setting current parameter values to initial parameter values. An optimization method is performed which iteratively minimizes the loss function. During each iteration, a steepest direction of the loss function is calculated by determining the gradient of the loss function at the current parameter values. A batch of samples included in training samples is selected. A matrix-free CG solver is applied to obtain an inexact solution to a linear system defined by the steepest direction of the loss function and a stochastic Hessian matrix with respect to the batch of samples. A descent direction is determined, and the parameter values are updated based on the descent direction. Following the optimization method, the parameter values are stored in relationship to the deep neural network.

Switching From Calendar-Based To Predictive Maintenance: A Leaner And Faster Software-Based Solution Orchestrating Data-Driven Forecasting Models

View page
US Patent:
20190188581, Jun 20, 2019
Filed:
Dec 18, 2017
Appl. No.:
15/844728
Inventors:
- Munich, DE
Ioannis Akrotirianakis - Princeton NJ, US
Amit Chakraborty - East Windsor NJ, US
International Classification:
G06N 5/04
G06N 99/00
Abstract:
A computer-implemented method for performing predictive maintenance includes executing a fleet prediction process. During this fleet prediction process, a plurality of fleet data records is collected. Each fleet data record comprises sensor data from a particular physical component in a fleet of physical components. A plurality of component maintenance predictions related to the fleet of physical components is generated. Each component maintenance prediction corresponds to a particular physical component. The plurality of component predictions are merged into one or more fleet maintenance predictions and the fleet maintenance predictions are presented to one or more users. Following the fleet prediction process, a next execution of the fleet prediction process is scheduled based on the fleet maintenance predictions.

Dimensionality Reduction In Bayesian Optimization Using Stacked Autoencoders

View page
US Patent:
20190034802, Jan 31, 2019
Filed:
Jul 28, 2017
Appl. No.:
15/662917
Inventors:
- Munich, DE
Ioannis Akrotirianakis - Princeton NJ, US
Amit Chakraborty - East Windsor NJ, US
International Classification:
G06N 3/08
G06F 17/11
Abstract:
The present embodiments relate to reducing the input dimensions to a machine-based Bayesian Optimization using stacked autoencoders. By way of introduction, the present embodiments described below include apparatuses and methods for pre-processing a digital input to a machine-based Bayesian Optimization to a lower the dimensional space of the input, thereby lowering the bounds of the Bayesian optimization. The output of the Bayesian Optimization is then projected back into the original dimensional space to determine input and output values in the original dimensional apace. As such, the optimization is performed by the machine in a lower dimension using the stacked autoencoder to constrain the input dimensions to the optimization.

Efficient Calculations Of Negative Curvature In A Hessian Free Deep Learning Framework

View page
US Patent:
20180101766, Apr 12, 2018
Filed:
Oct 11, 2016
Appl. No.:
15/290154
Inventors:
- Munich, DE
Ioannis Akrotirianakis - Princeton NJ, US
Amit Chakraborty - East Windsor NJ, US
International Classification:
G06N 3/08
Abstract:
A method for training a deep learning network includes defining a loss function corresponding to the network. Training samples are received and current parameter values are set to initial parameter values. Then, a computing platform is used to perform an optimization method which iteratively minimizes the loss function. Each iteration comprises the following steps. An eigCG solver is applied to determine a descent direction by minimizing a local approximated quadratic model of the loss function with respect to current parameter values and the training dataset. An approximate leftmost eigenvector and eigenvalue is determined while solving the Newton system. The approximate leftmost eigenvector is used as negative curvature direction to prevent the optimization method from converging to saddle points. Curvilinear and adaptive line-searches are used to guide the optimization method to a local minimum. At the end of the iteration, the current parameter values are updated based on the descent direction.
Control profile
Ioannis Akrotirianakis from Princeton, NJ, age ~56 Get Report