Automatic synchronization of cloned directed acyclic graph
A directed acyclic graph, asynchronous technology, applied in the direction of biological neural network models, probabilistic networks, and other database retrieval, can solve problems such as limited flexibility
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment Construction
[0034] Building, training, and utilizing neural networks that exceed the computational and / or memory capacity of a single machine is complex. Not only does the processing on different worker nodes need to be precisely synchronized, but there are also important and complex communication patterns between the different nodes.
[0035] Implementing such neural networks with existing frameworks and tools is difficult, time-consuming, and brittle, and even small changes to the network architecture require substantial changes to the underlying code. Also, after training a neural network, there needs to be a mechanism to load the weights and use them to make predictions on new data, but often to make predictions on a different number of machines and / or a different input size than was used to train the model. The current approach would require significant overhauls of the code used for training to support this use case.
[0036] The present disclosure relates to systems and methods fo...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 


