The last few years have witnessed a tremendous growth of the demand for wireless services and a significant increase of the number of mobile subscribers. A recent data traffic forecast from Cisco reported that the global mobile data traffic reached 1.2 zettabytes per year in 2016, and the global IP traffic will increase nearly threefold over the next 5 years. Based on these predictions, a 127-fold increase of the IP traffic is expected from 2005 to 2021. It is also anticipated that the mobile data traffic will reach 3.3 zettabytes per year by 2021, and that the number of mobile-connected devices will reach 3.5 per capita.
With such demands for higher data rates and for better quality of service (QoS), fifth generation (5G) standardization initiatives, whose initial phase was specified in June 2018 under the umbrella of Long Term Evolution (LTE) Release 15, have been under vibrant investigation. In particular, the International Telecommunication Union (ITU) has identified three usage scenarios (service categories) for 5G wireless networks: (i) enhanced mobile broadband (eMBB), (ii) ultra-reliable and low latency communications (uRLLC), and (iii) massive machine type communications (mMTC). The vast variety of applications for beyond 5G wireless networks has motivated the necessity of novel and more flexible physical layer (PHY) technologies, which are capable of providing higher spectral and energy efficiencies, as well as reduced transceiver implementations.
1. IEEE Signal Processing Magazine
2. Signal Processing Digital Library*
3. Inside Signal Processing Newsletter
4. SPS Resource Center
5. Career advancement & recognition
6. Discounts on conferences and publications
7. Professional networking
8. Communities for students, young professionals, and women
9. Volunteer opportunities
10. Coming soon! PDH/CEU credits
Click here to learn more.
Sequence generation tasks, such as neural machine translation (NMT) and abstractive summarization, usually suffer from exposure bias as well as the error propagation problem due to the autoregressive training and generation. Many previous works have discussed the relationship between error propagation and the accuracy drop problem (i.e., the right part of the generated sentence is often worse than its left part in left-to-right decoding models). In this paper, taking NMT as a typical sequence generation task, we measure the accuracy of the generated sentence with various metrics and conduct a series of analyses to deeply understand the accuracy drop problem. We obtain several interesting findings. First, The role of error propagation on accuracy drop is overstated in the literature, although it is indeed a cause to the accuracy drop problem. Second, Characteristics of a language play a more important role in causing the accuracy drop problem: the left part of the generated sentence in a right-branching language (e.g., English) is more likely to be more accurate than its right part, while the right part is more accurate for a left-branching language (e.g., Japanese). Our discoveries are also confirmed on other generation tasks (e.g., image captioning, abstractive summarization and language modeling) with multiple left/right-branching languages, as well as in various model structures.
© Copyright 2019 IEEE – All rights reserved. Use of this website signifies your agreement to the IEEE Terms and Conditions.
A not-for-profit organization, IEEE is the world's largest technical professional organization dedicated to advancing technology for the benefit of humanity.