An application helps you to run many experiments with different parameters in parallel. You just need to feed into application your experiment flow and parameters (using a web service) and you can track the results of your experiment using a unique identifier will be provided by the application
just calling a post service like the following:
POST /alexperiment/start_alexperiment HTTP/1.1
Host: localhost:8000
Authorization: ApiKey admin:ABC123
Content-Type: application/json
Cache-Control: no-cache
Postman-Token: 1deb1b8c-9c8e-4117-b2eb-8f02f3a45de8
{
"workflows": [{
"DataPhase": [{
"task_name": "Data",
"parameters": [{
"name": "path",
"possible_values": ["/home/mohamedsamir/Alexperiment/Data/Mylyn--CK.csv"]
}]
}],
"ClassificationPhase": [{
"task_name": "DNN",
"parameters": [{
"name": "nb_epoch",
"possible_values": ["100"]
}, {
"name": "batch_size",
"possible_values": ["10"]
}, {
"name": "n_splits",
"possible_values": ["10"]
}, {
"name": "scoring",
"possible_values": ["roc_auc", "accuracy"]
}]
}]
}]
}
will initiate a new run for you and an id will be provided to get the results of your run. The response of this service will be like the following:
{
"run_id": 18
}
To get the results of your run you can call the following GET service:
GET /alexperiment/run_results?id=18 HTTP/1.1 Host: localhost:9002 Authorization: ApiKey admin:ABC123 Content-Type: application/json Cache-Control: no-cache Postman-Token: ecb48c20-a480-4625-9f30-d2e9854cf1d8
The output of this service will be something like this:
{
"results": [
[
{
"description": {
"DataPhase": {
"Data": {
"path": "/home/mohamedsamir/Alexperiment/Data/Mylyn--CK.csv"
}
},
"ClassificationPhase": {
"DNN": {
"nb_epoch": "100",
"batch_size": "10",
"n_splits": "10",
"scoring": "roc_auc"
}
}
},
"results": {
"MEAN": "0.4273460688086292",
"STD": "0.0661907786255272"
}
},
{
"description": {
"DataPhase": {
"Data": {
"path": "/home/mohamedsamir/Alexperiment/Data/Mylyn--CK.csv"
}
},
"ClassificationPhase": {
"DNN": {
"nb_epoch": "100",
"batch_size": "10",
"n_splits": "10",
"scoring": "accuracy"
}
}
},
"results": {
"MEAN": "0.79163271776175",
"STD": "0.051743500958311144"
}
}
]
]
}
| License: | MIT |
|---|