Launch Basics

To launch an experiment only one file is needed and possibly a second file.

Experiment parameters

To initialize any experiment detail is needed on what experiment to launch and what parameters are needed. This file is called init.yaml and follows the YAML file format.

• Documentation on parameters to include in init.yaml can be found at http://[next-url]:8000/assistant/doc/<app_id>/pretty
  • where <app_id> is one of the apps (i.e., PoolBasedTripletMDS).
  • The links to these docs are available from http://[next-url]:8000/home
• Example of various init.yaml file can be found in examples/.

One example is here:

app_id: PoolBasedTripletMDS
args:
  alg_list:
    - {alg_id: CrowdKernel, alg_label: Test, test_alg_label: Test}
    - {alg_id: STE, alg_label: STE, test_alg_label: Test}
    - {alg_id: RandomSampling, alg_label: RandomSampling, test_alg_label: Test}
    - {alg_id: UncertaintySampling, alg_label: UncertaintySampling, test_alg_label: Test}
  algorithm_management_settings:
    mode: fixed_proportions
    params:
    - {alg_label: Test, proportion: 0.25}
    - {alg_label: STE, proportion: 0.25}
    - {alg_label: RandomSampling, proportion: 0.25}
    - {alg_label: UncertaintySampling, proportion: 0.25}
  d: 2
  debrief: Test debrief
  failure_probability: 0.01
  instructions: Test instructions
  participant_to_algorithm_management: one_to_many

Targets

The second file is a targets zip file. Every file in this ZIP file is uploaded to S3 and allows the user to see images you specify.

This ZIP file can also a single file, either a

• CSV file. One line per textual target.
• JSON file. This element is parsed and passed in as the targets parameter in args.

If you upload to S3, you need to provide an AWS Bucket Name, AWS Key ID and an AWS Secret Access Key.