You are not logged in. Click here to log in.

Application Lifecycle Management

Search In Project

Search inClear

ann04-select#39050/HEAD / v7
Tags:  not added yet

ann04-select[SERVICE-39050]

Tracker: Services Priority: NormalNormal Status: New
Submitted by: odavidJan 08 2018 11:30 Modified by: sidereusJan 23 2018 16:19 Type:
Model
Path: m/select/1.0 Context: csip-ann Version: 1.0
Endpoints: --
Maintainers:
Website: -- Apps:
--
Test Status:
--
Last Tested: -- Test Count: --
Test Success Rate: -- Failed Tests: --
Traceability
Loading…
Description

Selection of the more accurate ANNs

After n-runs of the training service, the most accurate ANNs are selected based on goodness of validation errors. User is asked to set 3 parameters in the payload (defaults provided):

  1. mechanism (default percentile): the selection algorithm;
  2. threshold (default 95): the threshold;
  3. error (default nashSutcliffe): the error on which selecting the best ANNs.

Available selection algorithms

Following the list of available selection algorithm:

  1. percentile: select the ANNs that outperform with respect to a threshold error evaluated as a result of provided percentile from probability distribution of errors;
  2. number: select the best n-ANNs (n- is a user-provided number);
  3. error: select the best ANNs with respect to a user-provided threshold error.

Old ANNs invalidation

Every time the RAW collection is updated with new data,
values_id
changes. This invalidates the neural networks already trained because the select service checks if the already trained network has the same
values_id
that is in RAW.

DirectedAcyclicGraph plugin failed: Could not find plugin DirectedAcyclicGraph
(com.ecyrd.jspwiki.plugin.PluginException:Could not find plugin DirectedAcyclicGraph, java.lang.ClassNotFoundException:Class 'DirectedAcyclicGraph' not found in search path!)

{
    "metainfo": {},
    "parameter": [
        {
            "name": "annName",
            "value": "test_ann"
        },
        {
            "name": "mechanism",
            "value": "percentile"
        },
        {
            "name": "threshold",
            "value": 95
        },
        {
            "name": "error",
            "value": "nashSutcliffe"
        }
    ]
}
{
    "metainfo": {
        "status": "Finished",
        "suid": "d415ae19-f5d1-11e7-9386-8b11ee17256d",
        "cloud_node": "10.1.33.2",
        "request_ip": "129.82.23.199",
        "service_url": "http://csip.engr.colostate.edu:8088/csip-ann/m/select/1.0",
        "csip.version": "$version: 2.2.4 3c276d66bde6 2017-12-15 od, built at 2018-01-09 23:43 by jenkins$",
        "tstamp": "2018-01-09 23:45:21",
        "cpu_time": 834,
        "expiration_date": "2018-01-09 23:45:52"
    },
    "parameter": [
        {
            "name": "annName",
            "value": "test_ann"
        },
        {
            "name": "mechanism",
            "value": "percentile"
        },
        {
            "name": "threshold",
            "value": 95
        },
        {
            "name": "error",
            "value": "nashSutcliffe"
        }
    ],
    "result": [
        {
            "name": "status",
            "value": "ok"
        },
        {
            "name": "error",
            "value": "nashSutcliffe"
        },
        {
            "name": "error treshold",
            "value": 0.8259409682325399
        },
        {
            "name": "selected anns",
            "value": 15
        }
    ]
}
Details
Comments & Attachments
Associations (1)
Children
SCM Commits
History (7)
Baselines
All (7)