Dynamic reconfiguration training computer architecture
A dynamic reconfiguration training machine learning computer architecture is disclosed. According to some aspects, a computing machine accesses a configuration file. The configuration file specifies parameters for a machine learning session. The computing machine trains a machine learning module to solve a problem, where the machine learning module operates according to the parameters specified in the configuration file. The computing machine generates an output representing the trained machine learning module.
1. A machine learning apparatus, the apparatus comprising:
processing circuitry and memory; the processing circuitry to:
access a configuration file, the configuration file specifying parameters for a machine learning session, the configuration file having a tree structure, the tree structure comprising branches that separate the parameters into categories including one or more of: staging parameters, data ingest parameters, neural network parameters, driver parameters, and reporting parameters;
train a machine learning module to solve a problem, wherein the machine learning module operates according to the parameters specified in the configuration file; and
generate an output representing the trained machine learning module.
2. The apparatus of claim 1 , the processing circuitry further to:
execute the trained machine learning module to solve the problem.
3. The apparatus of claim 1 , wherein the configuration file is a JSON (JavaScript Object Notation) file.
4. The apparatus of claim 1 , wherein the staging parameters comprise a SQL query for data extraction and a database path.
5. The apparatus of claim 1 , wherein the data ingest parameters comprise training data parameters, a training batch size, and a data ingest size.
6. The apparatus of claim 1 , wherein the neural network parameters comprise a neural network architecture parameter, a kernel parameter, a convolution pool parameter, a weight decay, and a number of layers.
7. The apparatus of claim 1 , wherein the driver parameters comprise a maximum number of training epochs, a logging frequency for the training epochs, and a driver type.
8. The apparatus of claim 1 , wherein the reporting parameters comprise a window size, a mask value, and a step size.
9. A non-transitory machine-readable medium for machine learning, the machine-readable medium storing instructions which, when executed by processing circuitry of one or more machines, cause the processing circuitry to:
access a configuration file, the configuration file specifying parameters for a machine learning session, the configuration file having a tree structure, the tree structure comprising branches that separate the parameters into categories including one or more of: staging parameters, data ingest parameters, neural network parameters, driver parameters, and reporting parameters;
train a machine learning module to solve a problem, wherein the machine learning module operates according to the parameters specified in the configuration file; and
generate an output representing the trained machine learning module.
10. The non-transistory machine-readable medium of claim 9 , the processing circuitry further to:
execute the trained machine learning module to solve the problem.
11. The non-transistory machine-readable medium of claim 9 , wherein the configuration file is a JSON (JavaSctipt Object Notation) file.
12. The non-transistory machine-readable medium of claim 9 , wherein the staging parameters comprise a SQL query for data extraction and a database path.
13. The non-transistory machine-readable medium of claim 9 , wherein the data ingest parameters comprise training data parameters, a training batch size, and a data ingest size.
14. The non-transistory machine-readable medium of claim 9 , wherein the neural network parameters comprise a neural network architecture parameter, a kernel parameter, a convolution pool parameter, a weight decay, and a number of layers.
15. The non-transistory machine-readable medium of claim 9 , wherein the driver parameters comprise a maximum number of training epochs, a logging frequency for the training epochs, and a driver type.
16. The non-transistory machine-readable medium of claim 9 , wherein the reporting parameters comprise a window size, a mask value, and a step size.
17. A machine learning method comprising:
accessing a configuration file, the configuration file specifying parameters for a machine learning session, the configuration file having a tree structure, the tree structure comprising branches that separate the parameters into categories including one or more of: staging parameters, data ingest parameters, neural network parameters, driver parameters, and reporting parameters:
training a machine learning module to solve a problem, wherein the machine learning module operates according to the parameters specified in the configuration file; and
generate an output representing the trained machine learning module.
18. The method of claim 17 , wherein the configuration file is a JSON (JavaScript Object Notation) file.