Parameters models

:tune

Tune a model from io data.

• Default value: nothing
• Accepted: nothing
• Method argument: model
• Example: model(:tune, project_name = "project", model_name = "model")

Mandatory parameters

The following parameters are mandatory in order to tune a model from input-output data.

project_name

Set project name of the tuning machine. After computation the model will be saved in this project

• Default value: nothing
• Accepted: string
• Method argument: model(:tune, ...)
• Example: model(:tune, project_name = "project")

io

Set the input-output data for tuning a model.

• Default value: nothing
• Accepted: string
• Method argument: model(:tune, ...)
• Example: model(:tune, io = "io_name")

Optional parameters

Model parameters

model_architecture

Set architecture of the tuning model.

• Default value: fnn
• Accepted: string: linear, fnn, rbf, icnn, resnet, polynet, densenet, neuralnetodetype1, neuralnetodetype2, rnn, lstm, gru
• Method argument: model(:tune, ...)
• Example: model(:tune, model_architecture = "resnet")

Please note that some architure allow iterative algorithms and hyperparameters are optimised, such as:

where, [ ] not iterative and [x] iterative.

model_exploration

Set the model exploration.

• Default value: ["fnn", "rbf", "icnn", "resnet", "polynet", "densenet"]
• Accepted: vector of string: ["fnn", "rbf", "icnn", "resnet", "polynet", "densenet", "neuralnetodetype1", "neulranetodetype2", "rnn", "lstm", "gru"]
• Method argument: model(:tune, ...)
• Example: model(:tune, model_exploration = ["fnn", "resnet"])

model_sample_time

Set the model sample time of the neulranetodetype2.

• Default value: 1.0 second
• Accepted: Float
• Method argument: model(:tune, ...)
• Example: model(:tune, model_sample_time = 5.0)

Computation parameters

computation_maximum_time

Set the training time of the model.

• Default value: 15 minutes
• Accepted: Dates
• Method argument: model(:tune, ...)
• Example: model(:tune, maximum_time = Dates.Minute(5))

computation_solver

Set the solver for tuning model.

• Default value: adam
• Accepted: string, adam, radam, oadam, nadam, lbfgs, oaccel, pso, lls
• Method argument: model(:tune, ...)
• Example: model(:tune, method_solver = "lbfgs")

Please note that some solvers are specific to model, such as:

where, [ ] depicts no supported solvers and [x] depicts supported solvers.

computation_processor

Set the processor used during computation.

• Default value: cpu_1
• Accepted: String, cpu1, cputhreads, cpu_processes
• Method argument: model(:tune, ...)
• Example: model(:tune, computation_processor = "cpu_threads")

computation_verbosity

Set the therminal verbosity during computation.

• Default value: 0
• Accepted: 0 (no information), 1 (fragmented information) or greather than 1 (full information)
• Method argument: model(:tune, ...)
• Example: model(:tune, computation_verbosity = 5)

Neural networks parameters

neuralnet_activation_function

Activation function for neural networks.

• Default value: relu
• Accepted: String: relu, sigmoid, swish, tanh
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_activation_function = "relu")

neuralnet_minimum_epochs

Minimum epochs hyperparameter value.

• Default value: 50
• Accepted: integer and lower than neuralnet_maximum_epochs
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_minimum_epochs = 50)

neuralnet_maximum_epochs

Maximum epochs hyperparameter value.

• Default value: 500
• Accepted: integer and higher than neuralnet_minimum_epochs
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_maximum_epochs = 500)

neuralnet_minimum_layers

Minimum hidden layers hyperparameter value.

• Default value: 1
• Accepted: integer and lower than neuralnet_maximum_layers
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_minimum_layers = 1)

neuralnet_maximum_layers

Maximum hidden layers hyperparameter value.

• Default value: 6
• Accepted: integer and higher than neuralnet_minimum_layers
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_minimum_layers = 6)

neuralnet_minimum_neuron

Minimum neuron hyperparameter value.

• Default value: 3
• Accepted: integer and lower than neuralnet_maximum_neuron
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_minimum_layers = 3)

neuralnet_maximum_neuron

Maximum neuron hyperparameter value.

• Default value: 10
• Accepted: integer higher than neuralnet_minimum_neuron
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_maximum_neuron = 10)

neuralnet_batch_size

Batch size value when training neural network.

• Default value: 512
• Accepted: 128, 256, 512, 1024, 2048
• Method argument: model(:tune, ...)
• Example: model(:tune, neuralnet_batch_size = 512)

:ls

List the models that are available for a project.

• Default value: nothing
• Accepted: nothing
• Method argument: model
• Example: model(:ls; project_name = "...")

Mandatory parameters

The following parameters are mandatories in order to list model from a project.

project_name

Project name to list model.

• Default value: nothing
• Accepted: string
• Method argument: model(:ls, ...)
• Example: model(:ls, project_name = "project#1")

:rm

Remove model for a project.

• Default value: nothing
• Accepted: nothing
• Method argument: model
• Example: model(:rm, project_name = "...", name = "...")

Mandatory parameters

The following parameters are mandatories in order to delete model from a project.

project_name

Project name of the model to delete.

• Default value: nothing
• Accepted: string
• Method argument: model(:rm, ...)
• Example: model(:rm, project_name = "project1")

model_name

Name of the model to delete.

• Default value: nothing
• Accepted: string
• Method argument: model(:rm, ...)
• Example: model(:rm, model_name = "model1")

:stats

Presents information about the tuning model.

• Default value: nothing
• Accepted: nothing
• Method argument: model
• Example: model(:stats, project_name = "...", model_name = "...")

Mandatory parameters

The following parameters are mandatories in order to provide information about the tuning model from a project.

project_name

Project name of the model.

• Default value: nothing
• Accepted: string
• Method argument: model(:stats, ...)
• Example: model(:stats, project_name = "project1")

model_name

Name of the model.

• Default value: nothing
• Accepted: string
• Method argument: model(:stats, ...)
• Example: model(:stats, model_name = "model1")