A computer architecture modeled after the human brain and designed to solve problems that human brains solve well, such as recognizing patterns and making predictions from past performance. Neural networks are composed of interconnected computer processors that calculate a number of weighted inputs to generate an output. For example, an output might be the approval or rejection of a credit application. This output would be based on several inputs, including the applicant's income, current debt, and credit history. Some of these inputs would count more than others; cumulatively, they would be compared to a threshold value that separates approvals from rejections. Neural networks "learn" to generate better outputs by adjusting the weights and thresholds applied to their inputs.