All Versions
87
Latest Version
Avg Release Cycle
35 days
Latest Release
-

Changelog History
Page 4

  • v2.2.0 Changes

    • Made Export to CSV and Import CSV operations more flexible.
    • Workspace copy-paste uses more readable the YAML instead of JSON. (You can still paste old JSON workspaces if you have any lying around.)
    • Hold SHIFT while dragging boxes to align them on a grid.
    • Nicer comment box presentation.
    • Added Transform and Derive column operations for easier table manipulation.
    • Added Import union of table snapshots operation to import multiple snapshots at the same time.
    • Added Save to snapshot and Compute inputs operations, mostly for automation purposes.
    • Bug fixes and performance improvements. Thanks for all bug reports!

  • v2.1.0 Changes

    • Upgraded to Apache Spark 2.2.0.
    • Added SQL box variants with up to 10 inputs. (The long-term plan is to have a single box that can take any number of inputs.)
    • Lynx-branded design applied to most of the user interface.
    • Users can now set SQL box summaries.
    • SQL boxes can optionally persist their outputs.
    • Train a decision tree classification model can use String attributes.
    • Navigate directories by clicking on the path segments on the splash page.
    • The new box metadata editor makes it possible to change the type of a box and fix broken custom box references.
    • Popups can be more flexibly resized.
    • Arrows can be deleted by clicking on them.
    • A number of minor bugfixes and improvements. Thanks for the feedback!

  • v2.0.0 Changes

    • Replaced LynxKite projects with more powerful and flexible workspaces. Workspaces allow you to express full workflows, structure your work, re-use components, and easily navigate your experiments.
    • LynxKite 2.0 is incompatible with the 1.x series in the sense that 1.x instances can only be upgraded to 2.0 with the loss of all LynxKite state. (External data, such as Parquet files on HDFS, would survive.)
    • The remote API has completely changed.
    • The list of operations has changed. Several new operations have been added, many have been renamed, and some have been deleted.

  • v1.14.1 Changes

    • Upgraded to Apache Spark 2.1.1 to get the fix for SPARK-19857.

  • v1.14.0 Changes

    • Added separate training operations for decision tree models (Train a decision tree classification model and Train a decision tree regression model), so now you are able to train a decision tree model, look at its parameters and use it to create predictions.
    • Visualization up to 10× faster in some cases.
    • Added Discard segmentation links operation.
    • Added Merge parallel segmentation links operation.

  • v1.13.1 Changes

    • Fixed the type of numeric values coming from Oracle databases.

  • v1.13.0 Changes

    • Visualization settings UI now matches the new attribute UI.
    • When you try to overwrite a table or view, you now get a confirmation prompt instead of an error.
    • Tab-completion for LynxKite operations works in Python command-line and in Jupyter notebooks.
    • You can now filter for position (Double, Double) attributes.
    • Added Segment by geographical proximity operation to segment vertices using Shapefiles and coordinate vertex attributes. Can be used e.g. to segment by geo regions or points of interests.
    • New "Sample graph by random walks" operation added. It can be used to make small smart samples that preserves the structure of the network to speed up model creation.
    • Edge and scalar attribute notes added or augmented.
    • Merge vertices by attributes operation keeps links with base project.

  • v1.12.1 Changes

    • Admins can backup LynxKite project data from UI.
    • Column types are displayed in SQL result box.
    • Added Lookup Region operation for finding locations inside shapefiles.

  • v1.12.0 Changes

    • Restore Hive support. (It was unintentionally disabled with the Spark upgrade in LynxKite 1.11.0.)
    • Remote API works on PizzaKite now.
    • Added a table and column explorer to the SQL box. Click on the "Tables" button in the SQL box to give it a try!
    • Upgraded to Apache Spark 2.1.0.
    • You can assign icons to attributes/segmentations/scalars. (As suggested by Marton Szel. :+1:)
    • New "Copy scalar from other project" operation added. For example it can be used to take a model which was trained in an other project.
    • Clicking on a table or view scrolls to the global SQL box and executes the select * query on it.
    • You can submit SQL queries with Ctrl-Enter.
    • New Remote API methods: project.copy(), project.global_table_name(), project.vertices_table(), project.edges_table().
    • Spark Monitor doesn't send alarm in case of long-running tasks.

    Ecosystem:

    • Replace run_task.sh and show_target.sh with a new unified tasks.sh.
    • Renamed from "Lynx Data Automation Framework" to "Lynx Enterprise".

  • v1.11.0 Changes

    LynxKite:

    • Upgraded to Apache Spark 2.0.2.
    • The Spark 2.x upgrade involved migration to the new Spark machine learning API. This will have a positive impact in the long run, but for now the effects are:
      • Logistic regression in the "Predict from vertex attributes" operation becomes a binary classifier.
      • "Train a k-means clustering model" no longer automatically scales the selected attributes. This is actually helpful if the attributes are on the same scale already, but otherwise you need to manually scale them now.
      • Previously trained k-means models will give wrong results because scaling has been removed. You will have to train new models.
    • Fixed all reported bugs in the new project UI.
    • Vertex and edge ID attribute values are only optionally unique in "Import vertex attributes" and "Import edge attributes" operations.
    • New "Split Edges" operation added. It can be used to create multiple copies of the same edge, based on a repetition attribute.
    • "Derive vertex attribute" and "Derive edge attribute" operations can be evaluated on all vertices / edges using undefined input attribute values.
    • "Derive vertex attribute" and "Derive edge attribute" operations can return vectors of doubles or vectors of strings.