Setup and create the database files. This will include the connection to a SQL Server, the validating/creation of the database, the creation of all the tables planned to be used, and a DAO class for the objects that may be used in the web service.
Some tables that will be required:
DataSet - Parent entity to track a group of data files as a training set
id
created_timestamp
last_updated_timestamp
name
DataFile - a file used in the training or validation
id
file_location - relative path the system file containing the data
file_extenstion - extension of the file
dataset_id - fk to DataSet
parent_data_file - fk to another DataFile. Intended for when an original data file is processed for training. e.g. distorted, or standardized to specific length and/or format
created_timestamp
last_updated_timestamp
training_flag - boolean for whether the file will be used in training
validation_flag - boolean/tinyint, if true, then it will be used in the validation process
NeuralNetwork - Table to track the creation of a neural network
id - primary identifier
created_timestamp
last_updated_timestamp
name - name of the neural network
file_location - relative path to the system .pt file containing save information of the neural network
model
optimizer
... other information regarding the NN that may be useful/interesting to track
NeuralNetworkTraining - Table to track a neural network's progress over time
id
neural_network_id - foreign key to a specific neural network
created_timestamp
last_updated_timestamp
dataset_id - fk to a dataset
execution_time - cpu or real time for training to finish
architecture - information regarding the machine that processed the training
loss_info - not sure what all the fields would be, but we want to track the score of the neural network after training
minimum_loss -
average_loss -
maximum_loss -
Setup and create the database files. This will include the connection to a SQL Server, the validating/creation of the database, the creation of all the tables planned to be used, and a DAO class for the objects that may be used in the web service.
Some tables that will be required:
DataSet - Parent entity to track a group of data files as a training set
id created_timestamp last_updated_timestamp name
DataFile - a file used in the training or validation
id file_location - relative path the system file containing the data file_extenstion - extension of the file dataset_id - fk to DataSet parent_data_file - fk to another DataFile. Intended for when an original data file is processed for training. e.g. distorted, or standardized to specific length and/or format created_timestamp last_updated_timestamp training_flag - boolean for whether the file will be used in training validation_flag - boolean/tinyint, if true, then it will be used in the validation process
NeuralNetwork - Table to track the creation of a neural network
id - primary identifier created_timestamp last_updated_timestamp name - name of the neural network file_location - relative path to the system .pt file containing save information of the neural network model optimizer ... other information regarding the NN that may be useful/interesting to track
NeuralNetworkTraining - Table to track a neural network's progress over time
id neural_network_id - foreign key to a specific neural network created_timestamp last_updated_timestamp dataset_id - fk to a dataset execution_time - cpu or real time for training to finish architecture - information regarding the machine that processed the training loss_info - not sure what all the fields would be, but we want to track the score of the neural network after training minimum_loss - average_loss - maximum_loss -