0000108: FIXED: overflows in neural networks under some compilers

Viewing Issue Advanced Details [ Jump to Notes ]

[ View Simple ] [ Issue History ] [ Print ]

Category

Severity

Reproducibility

Date Submitted

Last Update

0000108

[ALGLIB] Data analysis

minor

always

2009-08-22 20:20

2009-08-22 20:21

Reporter

SergeyB

View Status

public

Assigned To

SergeyB

Priority

normal

Resolution

fixed

Platform

Status

resolved

Projection

none

OS Version

ETA

none

Fixed in Version

2.1.0

Product Version

Target Version

Product Build

Summary

0000108: FIXED: overflows in neural networks under some compilers

Description

Neural networks in 2.0.1 use tanh as activation function. Some compilers (Delphi 7, for example) have buggy standard libraries that can't calculate TanH of large arguments because of overflow (although final result is "1"). So if neural network is provided with inputs that are too large, such network will crash during training. Strictly speaking, it is Delphi bug, but we'll fix it anyway :)

This problem is fixed by using special code for inputs that are larger than 100.

Steps To Reproduce

Additional Information

Programming language

Unspecified

Attached Files

Relationships

There are no notes attached to this issue.

Issue History
Date Modified	Username	Field	Change
2009-08-22 20:20	SergeyB	New Issue
2009-08-22 20:20	SergeyB	Status	new => assigned
2009-08-22 20:20	SergeyB	Assigned To	=> SergeyB
2009-08-22 20:20	SergeyB	Programming language	=> Unspecified
2009-08-22 20:20	SergeyB	Status	assigned => resolved
2009-08-22 20:20	SergeyB	Fixed in Version	=> [NOT RELEASED YET] 2.1.0 (to be released in August 2009)
2009-08-22 20:20	SergeyB	Resolution	open => fixed
2009-08-22 20:21	SergeyB	Summary	FIXED: neural networks can generate overflows => FIXED: overflows in neural networks under some compilers
2009-08-22 20:21	SergeyB	Description Updated

Mantis 1.1.6[^]