Even though many manufacturers integrate remanufacturing into existing business models, it should be noted that such efforts are usually accompanied by a major concern for cannibalization of new product sales from remanufactured products. To deal with this problem, many manufacturers, such as Dell, adopt a “two-roof policy” where the sale of new products takes place in a store and their remanufactured products in another. However, in contrast, some manufacturers, including Apple and HP, adopt a “one-roof policy”, by which all new and remanufactured products are sold through one store/chain. Although the literature on remanufacturing has extensively addressed sustainability issues within operations management, little attention has been paid to how “differentiated roof policy” for the marketing of remanufactured products affects sustainability issues. To fill this gap, in this paper, the authors develop two theoretical models in which manufacturers have the flexibility to distribute new and remanufactured products (1) through a one-roof policy (Model O) or (2) through a two-roof policy (Model T), respectively, and strive to address the question of how differentiated roof policies impact sustainability issues related to remanufacturing operations. Among other results, the central result suggests that, if the manufacturers care about economic performance, distributing both products through a two-roof policy is an advantageous strategy. Conversely, if they care about environmental sustainability, one roof is the preferred strategy.
The loss function is a criterion to evaluate the learning quality of a deep convolutional neural network, which represents the gap between prediction and ground truth. However, as the most commonly used loss function in image classification tasks, Cross‐Entropy loss does not encourage the model to distinguish the similarity between features. In this work, the authors investigate inter‐class separability of similar features learnt by convolutional networks and propose a loss function called Error Refactor Loss (ER‐Loss). ER‐Loss is based on the error caused by convolutional networks; it can improve the inter‐class separability and is simple to implement and can easily replace the Cross‐Entropy loss. Compared with softmax loss, ER‐Loss adds a dynamic penalty item which can help ER‐Loss monitor the actual situation of model training and adjust the value of the penalty item according to model training. The ER‐Loss on CIFAR100 and part of ImageNet ILSVRC 2012 is evaluated and the experimental result showed that the ER‐Loss can improve the accuracy of the model.