18464996. NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING simplified abstract (QUALCOMM Incorporated)
Contents
- 1 NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Original Abstract Submitted
NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING
Organization Name
Inventor(s)
Wonseok Jeon of San Diego CA (US)
Mukul Gagrani of Milpitas CA (US)
Weiliang Zeng of San Diego CA (US)
Edward Teague of San Diego CA (US)
Burak Bartan of San Jose CA (US)
Piero Zappi of La Jolla CA (US)
Christopher Lott of San Diego CA (US)
NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING - A simplified explanation of the abstract
This abstract first appeared for US patent application 18464996 titled 'NEURAL DIRECTED ACYCLIC GRAPH (DAG) SCHEDULING VIA ONE-SHOT PRIORITY SAMPLING
Simplified Explanation
The abstract describes a method for scheduling tasks in an artificial neural network based on node priorities sampled from a computation graph.
- Sampling of node priorities from a computation graph according to a priority sampling policy.
- Converting node priorities to a schedule using a list scheduling function.
- Associating each node with a processor from a group of processors to optimize makespan.
- Performing tasks in the artificial neural network based on the generated schedule.
Potential Applications
This technology can be applied in various fields such as:
- Optimization of task scheduling in artificial neural networks.
- Improving the efficiency and performance of neural network operations.
Problems Solved
- Efficient allocation of tasks to processors in artificial neural networks.
- Optimizing the overall performance of the neural network by reducing makespan.
Benefits
- Increased efficiency in task scheduling.
- Enhanced performance of artificial neural networks.
- Reduction in processing time and resource utilization.
Potential Commercial Applications
Optimized task scheduling in artificial neural networks can be beneficial for:
- Cloud computing services.
- Data centers.
- Machine learning applications.
Possible Prior Art
Prior art in task scheduling algorithms for parallel processing systems may exist, but specific examples are not provided in the abstract.
Unanswered Questions
How does this method compare to existing task scheduling algorithms in terms of efficiency and performance?
The abstract does not provide a direct comparison with existing algorithms, leaving the question of its superiority unanswered.
What impact does the scheduling policy have on the overall performance of the artificial neural network?
The abstract does not delve into the specific effects of the scheduling policy on the network's performance, leaving this aspect open for further exploration.
Original Abstract Submitted
A processor-implemented method includes sampling, according to a priority sampling policy, a set of node priorities from a computation graph. Each node priority of the set of node priorities may be associated with a respective node on the computation graph. Additionally, each node may represent an operation of a task performed by an artificial neural network. The method also includes converting, via a list scheduling function, the node priorities to a schedule that associates each node of the computation graph with a processor of a group of processors of a device associated with the artificial neural network, the schedule associated with a makespan. The method further includes performing the task in accordance with the schedule.
