NeuroLab is a tool that lets you build Neural Network indicators in Wealth-Lab using drag and drop. You can drag any number of indicators into the Neural Network's Input Layer, establish the architecture of the Hidden Layer(s), and specify the indicator to predict in the Output Layer (typically ROC, which is percentage change, or LogReturn).
You then select a DataSet, and train the Neural Network. When the training is completed, you have a new instance of the Neural Network indicator (NNPredictor) that you can use in any chart, Building Block, or C# Coded Strategy.
- Fixed loading trained Neural Networks that contain Swish or Leaky ReLU activation functions. For already corrupted NNs you will need to re-train and save again.
- Further fixes for loading Swish and Leaky ReLU networks.
- Fix: NeuroLab's blank help window
- NeuroLab training can now run against a single symbol or a DataSet.
- Rebuild required for some changes to the Genetic Evolver.
- Fixed issue that prevented saving of a new, untrained neural network.
- Minor fix to Save-As behavior.
- Fixed issue that showed zero evaluation results after training if the NN was not previously saved.
- Fixes to make NL work better in the Genetic Evolver.
- Initial WL8 release.
- Rebuild required due to change in base WL7 framework.
- Handle exception when parsing Neural Network.
- Change to adapt to some assembly reference changes in WL7.
- Build required to interface with some new charting changes in WL7 Build 28.
- Flag the NNP indicator as a lengthy calculation, so it doesn't get enabled by default in Indicator Profiler.
- Sort order is now retained when list results are copied to clipboard.
- Integrate NeuroLab's chart with the WL7 Linked Charting.
- Added multi-column sort support.
- You can now set the recommended learning rate in the HyperParameters page based on the selected Optimization Method.
- Established more realistic default values.