Remark: It is always wise to verify the
checksums provided by original authors after downloading
application files from the internet. This helps prevent downloading
modifies files that are infected with viruses and other malware.
Please note that even some well established download sites bundle
their downloadable software with adware that in some cases turns
much more harmful than just causing slight annoyance by showing
unwanted advertisement. If the original authors provide checksums
for their application files (such that above), check that actual
checksums of downloaded software match the provided checksums (there
are plenty of simple applications such as HashForm that
make this task really simple). It is also recommended to check
downloaded files with strong antivirus tools. I use the VirusTotal online
service for that, which checks the uploaded files by more than 50
antivirus solutions existent on the market.
The application requires the 64 bit .NET framework (version
4.0 is used in Neurapp 1.3) or its alternative implementation such
as Mono. One of these is
available for most modern computing platforms and are many times
pre-installed with the operating system. If not yet installed, you
will have to download and install the .NET framework (for Widows
operating systems) or Mono
(cross platform) before using the software.
There may be problems with using the software on some older
platforms, e.g. Windows XP. For more details, see the System
Requirements Section at the IGLib pages. Currently, only the
version 1.1 is available for 32-bit systems.
Examples and Screenshots
Approximation of function of 1 variable.
Generating training data by sampling a user defined reference
function of two variables.
Neural network - based model of a user defined function of two
variables.Original function is shown for comparison, with errors
shown in color scale.
Zoomed-in detail of an area with larger approximation errors.
Plot of approximated data with contours shown.
There are a number of display options that make inspection of the
approximation and its errors really easy and clear. One can
individually switch on or off display of training points, grid,
surfaces and contours of the original function and its
approximation. Transparency of both surfaces can be continuously
adjusted. In this figure, both original and approximated surface
were drawn relatively opaque, therefore it is immediately clear
where approximated function has higher or lower values than the
original one. Contours of the original function are displayed, which
gives better feeling of 3D shapes.
Some publically available software based on this library includes:
- application for calculation and verification of checksums
(MD5, SHA-1, SHA-256, SHA-512) for files and text.
- a shell application with various cryptography-related
functionality (calculation of hash functions / checksums,
symmetric and asymmetric encryption, certificate management,
– an educational application for visually exploring features
of function approximation
with artificial neural networks (ANN). It creates 1D and 2D
ANN models of user
defined functions and provides visualization capabilities to
models with originals. It can be downloaded
software for exploring
multidimensional ANN-based models.