A DNN inference acceleration algorithm combining model partition and task allocation in heterogeneous edge computing system

Author(s):  
Lei Shi ◽  
Zhigang Xu ◽  
Yabo Sun ◽  
Yi Shi ◽  
Yuqi Fan ◽  
...  
2019 ◽  
Vol 6 (3) ◽  
pp. 5853-5863 ◽  
Author(s):  
Song Yang ◽  
Fan Li ◽  
Meng Shen ◽  
Xu Chen ◽  
Xiaoming Fu ◽  
...  

IEEE Access ◽  
2021 ◽  
Vol 9 ◽  
pp. 138200-138208
Author(s):  
Ping-Chun Huang ◽  
Tai-Lin Chin ◽  
Tzu-Yi Chuang

Author(s):  
Phanish Puranam

Division of labor involves task division and task allocation. An extremely important consequence of task division and allocation is the creation of interdependence between agents. In fact, division of labor can be seen as a process that converts interdependence between tasks into interdependence between agents. While there are many ways in which the task structure can be chunked and divided among agents, two important heuristic approaches involve division of labor by activity vs. object. I show that a choice between these two forms of division of labor only arises when the task structure is non-decomposable, but the product itself is decomposable. When the choice arises, a key criterion for selection between activity vs. object-based division of labor is the gain from specialization relative to the gain from customization.


Sign in / Sign up

Export Citation Format

Share Document