On the principle of empirical risk minimization based on averaging aggregation functions

Минимизации Робастных Оценок Сумм Параметризованных Функций

“…Для дальнейшего сокращения объема вычислений можно воспользоваться приемом, который был применен при построении алгоритма IRLS (iteratively reweighted least squares) для восстановления линейной регрессии при наличии выбросов. Так, в [17] была предложена процедура итеративно перевзвешенной минимизации эмпирического риска, которую назовем IR-ERM (iteratively reweighted empirical risk minimization). Она представляет собой итерационный процесс, в котором на каждом шаге t сначала вычисляются компоненты градиента v k = v k (w t ,z t ), а затем решается задача минимизации взвешенного эмпирического риска:…”

A Face Detection Method Based on Sliding Window and Support Vector Machine

“…Therefore, if we can find solutions to these problems and successfully construct a face detection and tracking system, it will provide important enlightenment for solving other similar complex pattern detection problems. Therefore, if we can find solutions to these problems and successfully construct a face detection and tracking system, it will provide important enlightenment for solving other similar complex pattern detection problems.Common methods for face recognition include face recognition based on set features [8], statistical principles [9], model-based methods [10], active shape module-based methods [11], active representation model-based methods [12], and neural network-based methods [13].The support vector machine method is based on the VC dimension theory with statistical learning theory [14] and the principle of structural risk minimization [15]. According to the limited sample information, it seeks the best compromise between the complexity of the model and the learning ability in order to obtain the best generalization ability.…”