Machine learning

Combining Generative and Discriminative Methods for Pixel Classification with Multi-Conditional Learning

“…Multi-conditional models seek a tradeoff between these two extremes. A thorough analysis of the benefits of multi-conditional training must therefore consider the triple tradeoff between model complexity, amount of training data and test accuracy, fundamentally inherent to all supervised machine learning problems [2].…”

“…In the Berkeley segmentation database 2 An example from the database is shown in Figure 1. The users' labels are specified by a tri-map obtained with a lasso or pen tool as shown in Figure 1(a).…”

Computational Social Science Fusion Analytics: Combining Machine-Based Methods with Explanatory Empiricism

“…Algorithm use is an essential element of big data analytics. In some circumstances, machines can learn from data, and benefit from the intersection of machine learning, artificial intelligence, and data processing methods [27]. What is important is whether we can make scientific discoveries, and whether we can learn new theories, create deeper explanations, and make more effective predictions.…”

Recursive automatic bias selection for classifier construction

“…Therefore to find an accurate hypothesis in a reasonable amount of time, learning algorithms employ a restricted hypothesis space bias and a preference ordering bias for hypotheses in that space (Dietterich, 1990). Empirical comparisons among algorithms illustrate that no single bias exists that is best for all learning tasks (Weiss & Kapouleas, 1989;Aha, Kibler & Albert, 1991;Shavlik, Mooney & Towell, 1991;Salzberg, 1991).…”