Measurement of attribute classification power based on derivative information

2010 ◽  
Author(s):  
Yanfeng Fan ◽  
Zhixiao Yang ◽  
Xingxing Cheng ◽  
Dexian Zhang
Keyword(s):  
Attribute Classification ◽  
Derivative Information

Using Noise to Compute Error Surfaces in Connectionist Networks: A Novel Means of Reducing Catastrophic Forgetting

2002 ◽  
Vol 14 (7) ◽  
pp. 1755-1769 ◽  
Author(s):  
Robert M. French ◽  
Nick Chater
Keyword(s):  
Feedforward Networks ◽  
First Derivative ◽  
Error Surface ◽  
Connectionist Networks ◽  
New Information ◽  
Catastrophic Interference ◽  
Derivative Information

In error-driven distributed feedforward networks, new information typically interferes, sometimes severely, with previously learned information. We show how noise can be used to approximate the error surface of previously learned information. By combining this approximated error surface with the error surface associated with the new information to be learned, the network's retention of previously learned items can be improved and catastrophic interference significantly reduced. Further, we show that the noise-generated error surface is produced using only first-derivative information and without recourse to any explicit error information.

Multirate Integration for Multibody Dynamic Analysis With Decomposed Subsystems

1999 ◽  
Author(s):  
Sung-Soo Kim ◽  
Jeffrey S. Freeman
Keyword(s):  
Dynamic Analysis ◽  
Multibody System ◽  
Equations Of Motion ◽  
Low Frequency ◽  
Inertia Force ◽  
Integration Scheme ◽  
Integration Algorithm ◽  
Multibody Dynamic ◽  
Higher Order Derivative ◽  
Derivative Information

Abstract This paper details a constant stepsize, multirate integration scheme which has been proposed for multibody dynamic analysis. An Adams-Bashforth Moulton integration algorithm has been implemented, using the Nordsieck form to store internal integrator information, for multirate integration. A multibody system has been decomposed into several subsystems, treating inertia coupling effects of subsystem equations of motion as the inertia forces. To each subsystem, different rate Nordsieck form of Adams integrator has been applied to solve subsystem equations of motion. Higher order derivative information from the integrator provides approximation of inertia force computation in the decomposed subsystem equations of motion. To show the effectiveness of the scheme, simulations of a vehicle multibody system that consists of high frequency suspension motion and low frequency chassis motion have been carried out with different tire excitation forces. Efficiency of the proposed scheme has been also investigated.

scholarly journals A Study of the Stability of a Plate-Like Load Towed beneath a Helicopter

1971 ◽  
Vol 13 (5) ◽  
pp. 330-343 ◽  
Author(s):  
D. F. Sheldon
Keyword(s):  
Equations Of Motion ◽  
Physical Parameters ◽  
Recent Experience ◽  
Wind Tunnel Testing ◽  
Stability Derivatives ◽  
Direct Indication ◽  
Metric Dimensions ◽  
Set Up ◽  
The Stability ◽  
Derivative Information

Recent experience has shown that a plate-like load suspended beneath a helicopter moving in horizontal forward flight has unstable characteristics at both low and high forward speeds. These findings have prompted a theoretical analysis to determine the longitudinal and lateral dynamic stability of a suspended pallet. Only the longitudinal stability is considered here. Although it is strictly a non-linear problem, the usual assumptions have been made to obtain linearized equations of motion. The aerodynamic derivative data required for these equations have been obtained, where possible, for the appropriate ranges of Reynolds and Strouhal number by means of static and dynamic wind tunnel testing. The resulting stability equations (with full aerodynamic derivative information) have been set up and solved, on a digital computer, to give direct indication of a stable or unstable system for a combination of physical parameters. These results have indicated a longitudinal unstable mode for all practical forward speeds. Simultaneously the important stability derivatives were found for this instability and modifications were made subsequently in the suspension system to eliminate the instabilities in the longitudinal sense. Throughout this paper, all metric dimensions are given approximately.

The Symmetric Rank One Formula and its Application in Discrete Nonlinear Optimization

1988 ◽  
Author(s):  
J. Z. Cha ◽  
R. W. Mayne
Keyword(s):  
Hessian Matrix ◽  
Attractive Feature ◽  
Recursive Quadratic Programming ◽  
Rank One ◽  
Hereditary Properties ◽  
Symmetric Rank One ◽  
Derivative Information ◽  
Theoretical Considerations ◽  
Specific Search ◽  
Search Directions

Abstract The hereditary properties of the Symmetric Rank One (SRI) update formula for numerically accumulating second order derivative information are studied. The unique advantage of the SR1 formula is that it does not require specific search directions for development of the Hessian matrix. This is an attractive feature for optimization applications where arbitrary search directions may be necessary. This paper explores the use of the SR1 formula within a procedure based on recursive quadratic programming (RQP) for solving a class of mixed discrete constrained nonlinear programming (MDCNP) problems. Theoretical considerations are presented along with numerical examples which illustrate the procedure and the utility of SR1.

Sign in / Sign up

Export Citation Format

Share Document