Multi-dimensional Time-series Visualization and exploration

[thorwhalen]

• Data characteristics?
• Interface data format?
• Visualization features?

Data characteristics?

• Dimensionality: single & multiple --> Single can be particular case of multiple
• Regularity: regular & irregular
• Granularity: single vs batch

Interface data format?

• Dimensionality
  • Single: data_item := simple type (like int, float, bool, string,...)
  • Multiple: data_item := {channel_id: channel_data, ...}.
    • Later advanced interface: "The Matrix". (multi_stream[bt:tt, channels])
• Regularity
  • regular: needs a reference timestamp a sample rate
  • irregular: every data item needs a timestamp (e.g. (bt, data), ((bt, tt), data), {'bt': bt, 'data': data}, etc.)
• Granularity
  • single items
  • batches/chunks of data items (a.k.a slabs). Batches need timestamp or byte-offset

Visualization features?

• Statistics (e.g. duration etc.)
• Timeseries viz (e.g. spectrogram, volume graphs, model outputs, etc...)
• Playable (can hear the sound) (sometimes need to fiddle with sample rate (to make it audible))
• Cursor (to know visual location when played)
• Selectable: Be able to select intervals and do stuff with them (e.g. tag)
• Zoomable. (relevant: stock project (with SAP Hana))

Reference

Our stuff

• Creek design notes
• Multi Time Viz Components
• Search for "Resuable UI Components" in asana board

Extras

• A slang Session Scenario. Find there some screenshots of zoomable heatmaps, in the context of a user story.

[thorwhalen]

Relevant:

• Project starter
• Time-series serializers and deserializers