Product Tour

Download

Register

Buy

Add-Ons

Tips and Tricks

Online Tutorials

ANNI Forum

User Reviews

Links

About Us

Current Issues

FAQ

Home

Minor Maintenance and Bug Fixes

• Improved support for operating behind a proxy server
   
• Batch files tab remembers column location and width now
   
• Resolved problem with online information retrieval causing ANNI to lockup for slow connections
   
• Decreased file loading time caused by excessive auto-update attempts
   
• Resolved problem with Professional trial version not being accessible

Free Standard Version for paid Neural Science Prophet.Net users and Additional Graph Enhancements

ANNI Standard is now free to paid Neural Science Prophet.Net subscribers. Simply sign up for Prophet.Net services through ANNI and have unlimited use of ANNI Standard version as long as ANNI can sign into your Prophet.Net account when starting.

Large Graph has capability to show volume trend by color coding the volume bars as follows

• Green  = Up Day

• Red      = Down Day

• White   = No Change

Graph Enhancements and Minor Bug Fixes

Large Graph has capability to show the close values in line, bar or candlestick form. Added ZigZag trendline to allow easy viewing of the current trend. You can also hide the predicted values to allow better viewing of the close and trendline lines.

The following bugs have fixed in this version

• Error could occur during new file creation
   
• Volume wouldn't always show in large graph
   
• Retrieval of Business or Financial summaries were not always successful. Now ANNI will do multiple requests of the summaries until successful.

Professional Version Released

v2.20 signifies the first Professional release of ANNI. Read about the differences between the standard and the professional versions.

Minor Enhancements and Bug Fixes

• Added more detailed message on Configurations and Utilities that are not available during training or when a file is not open to prevent confusion.
   
• Added first time user wizard to assist in getting started.

Handles Recent Changes at Yahoo Finance

Yahoo Finance recently went through a change. In doing so, several web integration components broke. This update fixes all of the links and integration with Yahoo Finance.

Minor Enhancements and Bug Fixes

New toolbar button that allows reloading of the data - If you use another program to update your data, you no longer have to exit and reopen the ANNI file in order for it to reload the data file. Now by selecting this button (which is disabled during training, etc.) ANNI will force a data reload in order to pull in the updated data.

Buy Sell Signals Revamped

The Buy Sell Signals module has been completely rebuilt. The new systems design is superior to the old improving real world performance considerably. In addition a detailed report is available right on the Signals tab for easier analysis of the signals performance. An improved export of the signals results allows easy detailed analysis trade by trade.

Full MetaStock Data Support

ANNI has full direct MetaStock Data Support* for all versions of MetaStock Data (v7.0 and v8.0). This includes the newer XMASTER format that can hold more than 256 files using the .MWD data format.
* Although MetaStock Intraday data can be loaded into ANNI, the date display will not show the time intervals. In addition future predictions will be shown in daily intervals but technically represent whatever the time interval is in the data file.

Fixed Any Reported Bugs in v2.15

Minor Feature Changes

• Improved support for custom files that have reversed data entries
   
• Log Scaling capability for Large Graph

Minor Maintenance

• Fixed problem with Switching Files While Viewing Large Graph
   
• Fixed problem that could occur if a files data samples exceeded aprox 2000 entries and the maximum available entries where selected in the "Data Processing" view.

Minor Maintenance

Fixed problem with "Runtime Error" message occurring if using Auto-Run or Auto-Load features and if opening a file from within explorer (by double clicking a ANNI .nni file)
Balanced thread priority's in Buy/Sell Signals vs. Nets

Fundamental Data Updates Improved

The Fundamental Data updates have been separated into their own thread to prevent long file load times when updating Fundamental Data. This would cause ANNI to temporarily appear frozen. Now this is no longer the case.

Minor Maintenance

• Fixed "Moving Average Indicators Do Not Stay Selected" issue
   
• Fixed Dates not syncing with Buy Sell Signal Export vs. Graph
   
• Added Additional Support for Foreign Yahoo Finance Integration

Biggest Improvements ANNI has Ever Made!

New Learning Algorithm

Originally, ANNI's core learning algorithm was a compromise between speed and precision. Users have shown they are more concerned with precision than speed. Because of this, I've custom designed a learning algorithm like no other before it (see  Learning Algorithm ). This new algorithm is faster and better than the old one by far allowing a smoother transition of the weight space then the prior algorithm did. Another reason for this significant improvement is that the algorithm can scale itself independently to any problem domain (meaning each weight has it's own custom, dynamically scaled learning algorithm), the prior version did not scale itself independently. Having this algorithm deposited in each weight (not each neuron, or each net) gives superior non-linear separation of the search space like never before.

Additional Learning Enhancements

The momentum algorithm was changed to assure the new learning algorithm does not stop at a local minimum but continues until it comes as close as it can to a global minimum (another words - it can make it over a small hill if it needs to). And finally the weight decay algorithm was modified for the new learning algorithm in order to prevent excessively high weights (which tends to cause overtraining).

Enforced Diversity

ANNI in the past had a tendency to converge on one single input period if that input period had very strong signals. Code has been put in place that assures a greater diversity without affecting convergence. With the new version, it will be impossible for it to converge on one single input period (unless that's what the user sets it to do). This has improved ANNI's performance and generalization capabilities making the weighted average output more robust.

Improved Data Preprocessing

Version 1.0 had a basic scaling algorithm. Version 2.0 moved to a more dynamic version by analyzing the data and choosing the best scaling algorithm. This new version has added to that and allows combinations of scaling for each input used. So for example, volume can be logarithmically scaled while the close can be non-linear scaled. This has improved certain stocks performances considerably.

Professional Feature now In Standard

Prediction Range optimization has been moved out of the professional version and placed into the standard version. Now everyone will have the power of changing ANNI's prediction period (or allowing ANNI to optimize this for you). The prediction period is what ANNI trains towards (it's target value). So if the range was one day, then ANNI tries to predict tomorrows value. But if you change this to 5 days, then ANNI tries to predict 5 days from now. This sometimes helps ANNI find relationships easier and create systems that are more robust (i.e.. a mixed set of prediction ranges averaged together).

To Many More Moderate Changes to List - Download Now and Enjoy the New Power and Performance that has Made ANNI Famous.

Minor Maintenance

• Updated file links to "Auto Update" feature
   
• Fixed exiting problem with Win98 OS's
   
• Added support for Data Retriever companion program

Minor Maintenance

Various fixes to improve functionality and support for Win98SE and WinNT OS's

Changed Output Tab

The Output Tab now supports showing the Current Nets Output as before, or the Weighted Avg of the Best Nets. The settings are located in the Configurations/Program Settings dialog.

Minor Maintenance

Various fixes to improve functionality

Minor Maintenance

Various fixes to improve functionality

New Data Preprocessing - Improved Performance

With the current markets ups and downs, many stocks are reaching new lows (and some new highs). Previously ANNI has had trouble dealing with securities that are in this state since the values are in areas never seen or trained on before. With the new preprocessing of data, this problem has been severely reduced and in a lot of cases completely resolved. This new preprocessing technique also improves overall performance in correlation, value difference and training time by approximately 25%

New Web Browser Tab

With the addition of an optional Web Browser tab, online integration has never been better. The moment you open a file you will see it's current online chart available within ANNI itself. In addition you can have the Web Page auto refresh for you so the online chart is always up-to-date.

Data Column Selection Capability

Now you can select the columns you want to see in the Data tab. This includes any available indicator whether it is used in the neural nets or not. You can also set it to automatically select the columns the neural nets are training on. In addition ‘Daily Percent Change’, ‘Daily Price Change’ and the ‘Base Moving Average’ can be selected as well.

New Configuration Capabilities

Online Settings - Options for the new Web Browser tab plus
Advanced Training Options - Early Stopping Capability

Minor Feature Changes

• Auto updating feature while training for Advanced Views (Net's Weights, Input Period.)
   
• Charting capability of key performance values in the Genetic Info tab (values are remembered)
   
• Resizable graphs in tab views
   
• Graph Dialog, Error Correlation Chart and Advanced Views can be shown simultaneously while training or not (Modeless Dialog Boxes) in addition selection of a net within the main view will cause the related charts, views, etc. to change accordingly. This makes it much easier to navigate through the best nets via the main screen.

Minor Maintenance

Various fixes to improve functionality

Update to Handle Change in Yahoo Finance

Updated Data Retrieval module to handle change in data server

Minor Maintenance

Various fixes to improve functionality

Minor Feature Change

Genetic Info tab now shows progress of end of Generation Evaluation

Minor Maintenance

Various fixes to improve functionality

Improved Training

Slight changes to the learning algorithm to help speed up learning and improve performance

Added Feature to Input Analysis

Input Analysis can now be broken down into each separate data block or all data blocks. The 3 data blocks are the Training Entries, the Testing Entries and the Evaluation Entries.

File Size Decreased

Improved file serialization to decrease total file size making it easier to exchange files with other ANNI users. File size has bee decreased from a few KB to several MB depending on parameters used and current training progress.

Minor Maintenance

Various fixes to improve functionality

Improved Exporting Capabilities

You can export any available indicators with or without the neural networks core output in standard csv (comma separated values) format. Now you can use this data to import into your favorite spreadsheet program to have direct access to the indicators.

Improved Input Analysis

Keeping with the latest technology advances, ANNI has developed a much more advanced analysis of the affects of each of the inputs used to develop a network. It now fully takes into account the non-linear relationships created within a networks weights. This new technology has improved the input analysis's accuracy 100% while decreasing total time to analyze. ANNI will even show if an indicator has negative effect on the nets performance. This puts a very powerful tool into the users hands allowing them to quickly assess the value of the indicators used. Improving performance has never been easier.

Improved ASCII Custom Data File Support

ANNI can import your own custom ASCII data files. With or without headers. (ANNI will remember the custom format)  Simply assign your data to a particular column (i.e. high, low, open or volume) then use that column as an input indicator. The column used for close is the target value.

Improved Advanced Viewing Capabilities

All the Advanced Viewing Capabilities have been packaged up into one easy to use interface. In addition each has added functionality. This includes saving and printing of the graphs. And mouse position integration for specific data values in the graphs themselves.

Minor Maintenance

Various fixes to improve functionality

Added Zoom Capability to Large Graph

This gives the user the ability to Zoom in on a particular part of the Large Graph without having to enlarge the graph itself.

Minor Maintenance

Various fixes to improve functionality

Optimized Buy Sell Signals

ANNI creates optimized buy sell signals based on your investing style. These signals are optimized entirely separate from ANNI's core Neural Net technology so each module can focus on what it does best. Both modules can be trained simultaneously. The results can then be overlaid onto the Neural Networks outputs to give you an idea of were your next trade will occur. Results can be exported into a standard text file also.

Easier to Use

ANNI's results and performance are easier to understand more than ever before. The main screen has been changed to allow a quick assessment as to whether ANNI has converged onto stable patterns or not. All the neural net's performance are now laid out in a single tab with sortable columns for easy comparison to each other.

Fundamental Information

ANNI pulls in current fundamental information from the Internet and lays it out in an easy to understand format giving you clear signs of danger. Now your no longer limited to just technical analysis.