In recent years deep structured learning is practiced in many safety-critical applications. Deep learning algorithms are used in every field, such as face recognition (laptop, smartphones) and fingerprint recognition for financial applications. A deep neural network performs classification tasks. In designing any application using machine learning, enormous training data are required. Appropriate training gives the correct results. Other than training, many factors need to be considered viz labelling data: (manual task), expert advice, model selection, etc. In 2013 researchers observed that deep neural networks give expected results to the adversary by using adversarial samples during the testing or deployment phase without directly accessing the model, considering no knowledge about the model.
Adversary does use labels during the deployment phase, modify labels, and apply them to attack the victim model. Trustability of the deep neural network is a significant concern in many safety-critical applications, where input data contains noise and is not certain. Most frequently, uncertainty and noise in data are random, and adversary has incentives to change the data. As a preventive measure to detect malicious activities increases, adversaries become motivated to put extra effort into deceiving the algorithms. The security issue is then critical, especially for the applications which involve safety and monetary usage.
外文書商品之書封，為出版社提供之樣本。實際出貨商品，以出版社所提供之現有版本為主。部份書籍，因出版社供應狀況特殊，匯率將依實際狀況做調整。
無庫存之商品，在您完成訂單程序之後，將以空運的方式為你下單調貨。為了縮短等待的時間，建議您將外文書與其他商品分開下單，以獲得最快的取貨速度，平均調貨時間為1~2個月。
為了保護您的權益，「三民網路書店」提供會員七日商品鑑賞期(收到商品為起始日)。
若要辦理退貨，請在商品鑑賞期內寄回，且商品必須是全新狀態與完整包裝(商品、附件、發票、隨貨贈品等)否則恕不接受退貨。
