TinMan Technology Solves the Historic Issues with Artificial Neural Networks 

Historic Issue	TinMan Technology Resolution
Neural Networks are effective at pattern recognition (single step logic cycle) but little “actual thinking”	Modular architecture with limitless layers and simultaneous model-based processing states
Mathematics Intensive & difficult to implement/modify	Visual Design and Full abstraction of how and why neurons connect and model template approach. Assembly similar to actual thought.
Learning functions require narrow input ranges (e.g. 0-1 or -4 to 4) not “real-life”	Integrated fuzzy logic interpreters and flexible data input types.
Large input sets result in millions of feature vector permutations – long and not always successful learning cycles.	Modular architecture and training on a model-level basis – typically few inputs to permute to achieve full feature set variance.
Ability to learn and to generalize from limited input or missing data is key value, but requires supervision or alternative analysis to confirm agreeable outcome	Combination of fuzzy logic and ability to filter input set for output constraint eliminates undesired generalizations.
Significant computational requirements	State-relevant computation of modular approach only fires neurons related to the current task. Significant reduction in computation for learning and for autonomous operation.

TinMan technology fully abstracts and shields the user from the mathematical tedium associated with neural networks and machine learning technologies through a modular and templatized approach. Combined with optimized training algorithms and a  state of the art interface, TinMan provides highly productive and rapid development of a multi-layer, parallel processing, self-modifying neural network, ready for deployment in a dynamic environment .

The TinMan approach is a modular, interconnected, deep, multi-network system to provide speed and convenience throughout the design process, but also to eliminate historic limitations to applied artificial neural networks. TinMan AI systems are extremely well-suited for continuous processing of input data in a dynamic environment, allowing for the continuous, potentially perfect execution of all desired behaviors / outputs. However, AI systems developed in TinMan do not have to be autonomous (on their own and functioning with continuous stream of data) - they can be designed simply to compute one or more answers in response to a manual user event in the host application.

TinMan AI systems are "trained" (given knowledge on how to behave and the right choices to make under all conditions) through optimized learning algorithms that provide for speed and scale well beyond those of textbook algorithms for tuning networks. TinMan exposes these optimized training algorithms and the process of training a system in the form of a spreadsheet filled with auto-generated input scenarios, and the single click-and-assign process for directing desired behavior.