This is the NeuroDimension Newsletter, which you are receiving because you requested to stay informed about new developments at NeuroDimension. If you would like to stop receiving these newsletters, please see the bottom of this newsletter for instructions.
In this issue...
|
TradingSolutions v2.1 Build 031112 Now Available
TradingSolutions v2.1 Build 031112 is now available from the TradingSolutions website. This build adds several usability enhancements and improvements, as well as fixing several issues that were encountered in the previous builds of TradingSolutions v2.1.
|
|
Key improvements in this build include:
- Support for importing and streaming data from IQFeed.
- Support for importing end-of-day data from Quotes Plus.
- Improved status messages, data checking, and much more.
Download a free evaluation version of TradingSolutions or upgrade your software today at:
http://www.tradingsolutions.com
|
|
NeuroSolutions Technology Survey
The NeuroSolutions development team is currently focusing its efforts on adding new technologies to the software.
We have received many suggestions over time and have compiled a list of the most requested ones.
|
|
If you are interested in the future development of NeuroSolutions,
we would like to ask you the favor of filling out this short survey to let us know which types of neural network architectures you are most interested in.
For each one, please select one of the following three interest levels:
- Very Interested
- I would purchase/upgrade NeuroSolutions in order to build this type of network.
- Somewhat Interested
- I might experiment with this network type if it were included.
- Not Interested
- I would not use this network type.
|
Training Neural Networks for Control
Neural networks can be very useful in control systems. By "control systems", we mean the software which is used to control real-time processing in hardware or software. Examples of this include adjusting settings in a physical plant, tuning settings on hardware devices for different individuals, or purchasing supplies in response to changing prices or inventory conditions.
|
|
A Simple Approach
The easiest way to train and utilize a neural network for control is to collect a sufficient amount of data from the system, train the neural network, and then implement the system with fixed parameters (e.g. without the network learning while in operation). The data used for training a neural network should typically include multiple examples of the presence and absence of each condition being detected or requiring adjustment.
Adaptive Control
For control systems which need to adapt over time to new conditions, “online learning” may be needed. In other words, the training weights may need to be updated dynamically while the system is in use or “online”.
One way to approach this problem is to include the neural network training directly in the control application. As new data arrives, the neural network weights are continuously updated based on the data that arrives. This could be implemented by allowing NeuroSolutions to dynamically control the system while receiving live data feeds from the system, controlling NeuroSolutions through OLE automation, or by incorporating a learning neural network in the control application using Source Code Generation or the Custom Solution Wizard Developers level.
Safe Adaptation
|
Based on our experience, an intermediate approach based on both methods is often more practical and safe. In this approach, you can create a "batch" process that periodically sends a chunk of data to the neural network for training. After the new batch of data is trained, a validity check can be done against the new system. Once validated, the weights used by the controller can be updated.
|
|
|
Using this approach, the controller itself works only with the forward activation portion of the neural network (the easy/efficient part) while the training is done offline. The training can still be automated using one the methods described above, but it is generally under fewer time and processor constraints. This method can also be used to defer new training until certain criteria are met, such as when the performance of the system begins to degrade.
|
|
We appreciate your feedback! Please send us your comments or suggestions concerning this newsletter, our web site, or part of the NeuroDimension product line. Write to us at: newsletters@nd.com.
Have questions about NeuroDimension products? Send your questions to: info@nd.com
This issue and previous issues of this newsletter are available on the NeuroSolutions web site at: http://www.neurosolutions.com/newsletters.html
If you would prefer not to receive these newsletters or subsequent product updates from NeuroDimension, please reply to this letter with the subject heading changed to the word REMOVE.
Thank you again for your support of NeuroDimension products!
|
|
| |
