v0.4.0 - Support for Julia v1.0

Port to Julia 1.0

v0.3.5

This release changes the name of the package from StochasticSearch to NODAL, a recursive acronym for NODAL is an Open Distributed Autotuning Library.

This was done to better reflect the purpose and objectives of the package, which are to provide a set of tools for implementing parallel and distributed autotuners for different problem domains.

The examples and documentation were updated to use the new name.

v0.3.0

Major changes:

• Migrate to v0.5
• Add a Read The Docs documentation
• Implement Iterated Local Search
• Please, check the v0.3.0 roadmap for all improvements

All fixes and changes:

• Complete migration to v0.5 with no warnings.
• Add uniform search task initialization.
• Add the Iterated Local Search technique.
• Fix examples.
• Update documentation.
• Update logo.
• Add version shield to README

v0.2.2

Major changes:

• The package no longer supports Julia nightly due to breaking changes from Julia v0.4 to v0.5.
• The interface to Optim.jl was removed, since the two packages have very different objectives and the interface was not being used by any code, and could not be used with the latest search techniques.

List of all fixes and changes:

• Update README.md
• Update 'rosenbrock' example.
• Update parameter type in README.
• Remove functions left over from Optim interface
• Remove dependance on Optim.
• Remove unused variable in search techniques.
• Remove unused variable in search blocks.
• Fix #13.
• Add VERSION file.
• Remove support for Julia nightly.
• Add the 'Run' configuration type.
• Distribute RemoteRefs between workers.
• Add a new channel type.
• Change 'Result' constructor syntax.

v0.2.1

List of fixes and changes:

• Fix #6 (c200af2)
• Add a PermutationParameter (9a62e20)
• Techniques are independently initialized and configured (887d187)
• Techniques run separate remotecalls (351ef1a)
• Update constructor syntax (44cc51b and b488ff5)
• Using larger result Channels (7da389f)

v0.2.0: Goodbye 0.3.x

Changes to the API and Dictionaries made the package incompatible with v0.3.x.

v0.1.0: Build, Measure & Search

Features Implemented in v0.1.0:

• Building blocks for the implementation of common Stochastic Local Search algorithms:

  • First Improvement
  • Iterative First Improvement
  • Random Walk
  • Randomized First Improvement
  • Probabilistic Iterative Improvement
  • Greedy Construction
  • Iterative Greedy Construction
  • Simulated Annealing (rewrite with building blocks)

  (Check: Stochastic Local Search Algorithms: An Overview)

• Independent measurement module:

  • measure_mean!: Uses remotecall_fetch to get results from multiple workers.

• Independent search module.

• Tests pass in Julia stable (v0.3.x).

• Tests pass in Julia latest (v0.4).

Check the v0.1 Roadmap in this issue. The features intended for v0.1.x can be found here.

v0.0.5

Supports Julia v0.4 and v0.3.

Measurement and Searching are done in separate modules.

SLS building blocks for Iterative First Improvement.

v0.0.4

Supports Julia 0.3.11 and 0.4.

Search and Measurement (parallel) are still done in a single module.

v0.0.3

Adds experimental parallel evaluations.