WP4: Infrastructure
Space shortcuts
Page tree
Skip to end of metadata
Go to start of metadata
StatusCommentWhoComment
  •  
Fix broken APIMarco Soldati 
  •  
Create production_03 at 24hours cadenceMarco Soldati 
  •  
Create production_03 at 12h cadenceMarco Soldati 
  •  
Restart Property_Viewer for IoannisMarco Soldati 
  •  
Upgrade to PostgresQL 9.6Marco Soldati 
    
  •  
Create Jupyter Notebooks to access dataMarco SoldatiNot publicly available yet.
  •  

Update Verification Data Structure:

  • algorithm_run_id --> run_id
  • algorithm_name
  • output_type --> forecast_type
  • flaring_level
  • forecast_window
  • forecast_latency
  • base_issuing_time
  • property_set
  • start_date
  • end_date
  • data
  • algorithm_config_names, a list of [algorithm_config_name]
  • source_predictions, a list of [fc_id]
Marco Soldati 
  •  

Check data/data structure in predictions service

  • property_service has structure data/data (and contains properties of one active region)

  • prediction_service has structure data/prediction_data/data (prediction_data contains several predictions)

Marco Soldati 
  •  
Update Connexion framework to latest Swagger UI, if possible — not feasible at the momentMarco Soldati 
  •  
Create Tar file for postgres backupsMarco SoldatiRather trivial
  •  
Cleanup workflow_manager on msoldati accountMarco Soldati 
  •  
Fix broken prediction-viewerMarco SoldatiFixed by Colin, currently being improved by colin
  •  
Enable auto-updatesMarco SoldatiShould be straightforward
  •  
Check data gaps in 2013/2014Marco Soldati 
  •  
MonitoringMarco Soldati 
 Algorithm menu: the information-button message reads "Select "RandomForest" if you are unsure.". Is RandomForest the algorithm with the least gaps in results?  Manolis comment
 We would probably need a legend with a one-liner description of each prediction method.D. Shaun BloomfieldAccording to Manolis comments
 If it does not require too much effort, I would place an extra tab to show gaps of results over the period Jan 1, 2013 - Feb 1, 2018 According to Manolis comments
 Expert API: Moreover, what does the "Filter" option correspond to?–> Replace by "Text Filter"Colin KlauserAccording to Manolis comments
 There should be some reasoning as to the color of probability bars used. This could be coming out of validation. Is there any reasoning for this color at this point? According to Manolis comments
 The meaning of the "exclude flare history" option may be not clear to the user, is it about (not) taking into account the flare history of each AR? It is seems that of course (global) flare history must be used for learning, but this option seems to be about something else.D. Shaun BloomfieldProvide comments
 If "no results* is shown. Try to extract linkColin Klauser 
 

In
http://api.flarecast.eu/jupyter/prediction_data.html
there could be a first step for obtaining the fc_id from basic query parameters (dates...)

  
 

In
http://api.flarecast.eu/prediction/ui/
Retreive → Retrieve

There could be a link to some API documentation, for example I have a hard time finding which route to use to get a valid/good algorithm_config_name for http://api.flarecast.eu/prediction/analysis/gap_ratio

  
 WP2: Find out why midnight data is being lostD. Shaun Bloomfield 
 Rerun Flare_associsation on all data for a new dataset and compute the delta between the twoMarco Soldati 
 WP2: Look into Vector data and see why it performs so badly.D. Shaun Bloomfield 
 WP3: R vs Python code. Prediagnosis with the code.  
 

Full disk prediction service. Full disk at given timestamp. Full disk in time range.

Marco Soldati 

 

 

Low priority

StatusCommentWhoComment
  •  

SiRQL: for eq use IN operator rather than OR

  • Step 1: introduce 'in' operator in SIRQL for testing.  done
  • Step 2: use 'in' operator logic for 'eq' operator and remove 'in' operator.  – nice to have, but not much performance gain.
Marco Soldati

 

 

 

 

  • No labels