Amazon.com

About: Amazon.com is a company organization based out in Seattle, Washington, United States. It is known for research contribution in the topics: Computer science & Service (business). The organization has 13363 authors who have published 17317 publications receiving 266589 citations.

Hybrid machine/human computing arrangement

Abstract: A hybrid machine/human computing arrangement which advantageously involves humans to assist a computer to solve particular tasks, allowing the computer to solve the tasks more efficiently. In one embodiment, a computer system decomposes a task, such as, for example, image or speech comparison, into subtasks for human performance, and requests the performances. The computer system programmatically conveys the request to a central coordinating server of the hybrid machine/human computing arrangement, which in turn dispatches the subtasks to personal computers operated by the humans. The humans perform the subtasks and provide the results back to the server, which receives the responses, and generates a result for the task based at least in part on the results of the human performances.

Benefits of Reevaluating Real-Time Order Fulfillment Decisions

Abstract: When a customer orders online, an online retailer assigns the order to one or more of its warehouses and/or drop-shippers to minimize procurement and transportation costs based on the available current information. However, this assignment is necessarily myopic because it cannot account for any subsequent customer orders or future inventory replenishment. We examine the benefits of periodically reevaluating these real-time assignments. We construct near-optimal heuristics for the reassignment for a large set of customer orders to minimize the total number of shipments. Finally, we present evidence of significant saving opportunities by testing the heuristics on order data from a major online retailer.

Predictive reader power management

Abstract: Power management features of a reader device control an amount of electrical energy supplied to the reader device. The power management features include several power control categories that a user may select to affect an amount of power consumed by the reader device. One power control category controls power consumption based on a genre of an electronic book being processed by the reader device. Another power control category controls power consumption based on a time of day an electronic book is being processed by the reader device. Power control instructions that control how much power a reader device consumes may also be generated based on a characteristic of a user's interaction with the device.

TabTransformer: Tabular Data Modeling Using Contextual Embeddings

Abstract: We propose TabTransformer, a novel deep tabular data modeling architecture for supervised and semi-supervised learning. The TabTransformer is built upon self-attention based Transformers. The Transformer layers transform the embeddings of categorical features into robust contextual embeddings to achieve higher prediction accuracy. Through extensive experiments on fifteen publicly available datasets, we show that the TabTransformer outperforms the state-of-the-art deep learning methods for tabular data by at least 1.0% on mean AUC, and matches the performance of tree-based ensemble models. Furthermore, we demonstrate that the contextual embeddings learned from TabTransformer are highly robust against both missing and noisy data features, and provide better interpretability. Lastly, for the semi-supervised setting we develop an unsupervised pre-training procedure to learn data-driven contextual embeddings, resulting in an average 2.1% AUC lift over the state-of-the-art methods.

Unlearn Dataset Bias in Natural Language Inference by Fitting the Residual

Abstract: Statistical natural language inference (NLI) models are susceptible to learning dataset bias: superficial cues that happen to associate with the label on a particular dataset, but are not useful in general, e.g., negation words indicate contradiction. As exposed by several recent challenge datasets, these models perform poorly when such association is absent, e.g., predicting that “I love dogs.” contradicts “I don’t love cats.”. Our goal is to design learning algorithms that guard against known dataset bias. We formalize the concept of dataset bias under the framework of distribution shift and present a simple debiasing algorithm based on residual fitting, which we call DRiFt. We first learn a biased model that only uses features that are known to relate to dataset bias. Then, we train a debiased model that fits to the residual of the biased model, focusing on examples that cannot be predicted well by biased features only. We use DRiFt to train three high-performing NLI models on two benchmark datasets, SNLI and MNLI. Our debiased models achieve significant gains over baseline models on two challenge test sets, while maintaining reasonable performance on the original test sets.