Thinks

Windows application that provided high performance neural network solutions for use in real world business problems.

ThinksPro

The ThinksPro Windows Application released in 1996, provided better visualization tools, more algorithms and Dynamic Link Libraries so that the trained networks could be embedded in custom applications. ThinksPro was also the first environment to allow true reinforcement learning within user defined models.

ThinksPro Vision Toolkit

This new product is a toolkit for image understanding and recognition. It is designed to be the engine for any image recognition task in surveillance, industrial inspection, and scientific applications.

The ThinksProVision GUI is a java based (platform independent) application allowing viewing and labeling of images. The GUI includes a training set builder to generate training and test examples for training with ThinksPro 32 for the ThinksProVision Engine. The GUI allows the selection of "focus of attention" operators and classification methods, including Full Image classification and Sub Image classification for finding specific items in an image.

The ThinksProVision technology is based on a hierarchy of processing. The hierarchy mimics human vision in that the first step in processing is a "focus of attention" process which simplifies the subsequent processing. Our unique reduced connection neural network architectures for image processing insure the fastest possible processing times. Because it's a trainable technology, performance improvements can be easily realized by simply adding more data.

ThinksProVision Engine can be integrated into any platform environment including Windows and Linux.

Contact Logical Designs for further details related to this technology.

ThinksPro32 Toolkit

Currently, Logical Designs provides "leading edge" proprietary neural network technology as well as the best performing non-proprietary techniques available. With the release of ThinksPro32 in 1999, a complete Windows compliant application development environment, neural network technology can be developed and deployed in applications ranging from controllers to mainframes.

Logical Designs provides applications consulting, custom development (both hardware and software), and neural networks applications training and seminars.

Latest Version is 8.97 3/12/06

New Features 8.97

- Improved initialization of nets when missing VAR file.

New Features 8.96 and earlier

- Pass Through Analysis of inputs. This is similar to sensitivity analysis which indicates the importance of each input to the overall network performance. Pass Through Analysis does the sensitivity analysis output by output so that one can see for each output, what the contribution is of the various inputs.

- PolyTrn network growing technique. This is a proprietary technology for growing neural networks.

- Weights Analysis and weights pruning for PolyTrn.

- Added CALC(*) to allowed operation in VAR file (Oper=). The user can specific a more complex function of available inputs to be an input to a neural network.

- More error messages when loading data

- Correct handling of paths containing "." .

- Test while train for TP3.

- Sub Image Processing Element type for image processing

- Backward compatible for new parameter file variables.

- Changed Sensitivity Analysis Normalized Effect to ignore 0 effect variables.

- Output Class weight factors for better classification results.

- More Excel friendly .wlg and .sa files.