Version 1.0.1 - Bug fix

Fix a bug in the main algorithm where results could not be written in fits format.
Some general loop optimizations.

Version 1.0.0 - First production version!

This version is now production ready and has been validated as giving correct results.

Main features

• Main algorithm now used connected components between steps
• The Matrix class got significant changes and optimizations
• A few memory optimizations
• Solved a few bugs leading to wrong results in some cases

Future work

• Optimize the Fourier transform, too slow for now to be usable for small size pictures
• Add multi-picture bootstrapping capabilities

Version 0.6.0 - Bootstrap support enabled!

Project has been renamed to ALIAS and is now almost done.

Main features

• Main estimates are done and working independently
• Bootstrapping is now available for computation
• The main launcher now has a user-friendly interface
• The project no longer supports the external Eigen library
• Optimizations and bug fixes

Future work

• Complete support for Fourier convolution, currently in alpha stage
• Add MPI layer to allow distributed computing of bootstraps

Version 0.5.0 - First result without precomputed values

So many lambdas!

Normalization and regularization lambda values are now computed from Monte Carlo simulations directly.

Main features

• Monte Carlo simulations to compute normalization and regularization values
• Bug fixes, optimizations and general code improvement

Future work

Complete the estimation of the beta values

Version 0.4.1 - Bug fixes and performance improvement

Merge pull request #7 from philippeganz/dev

Dev

Version 0.4.0 - First working example on real data

AstroOperator is alive!

Release showing a working example on Chandra data. Result is computed 16 times faster and with 6 times less memory than MATLAB. Difference in result is 3%.

Main features

• Finalized the AstroOperator including tests
• Many improvements in the entire project

Future work

Complete the main algorithm

Version 0.3.0 - Operators done and working

Lots of operators!

Release to show the working operators.

Main features

• Addition of many operators:
  • Convolution
  • Abeltransform
  • Wavelet
  • Spline
  • Blur
• Many memory and multi-threading optimizations
• New folder structure to reflect class inheritance

Future work

Combine the operators into the main algorithm.

Version 0.2.0 - Class structure change

Class structure just got better!

Many improvements here and an important milestone for going on with the project.

Main features

• New LinearOp class as base class for Matrix/Operator class types
• DataContainer renamed to Matrix, meant to hold raw arrays of data
  • Added multiple matrix multiplication methods for better performance
  • Partial reshape of the class with many non-member functions added
  • Revisited many methods to make them make use of move semantics
• New Operator class, meant to apply a transformation to a Matrix
• Test suites now also provide performance calibration tools
• General improvement and bug fixes

Version 0.1.0 - Working example of FISTA

First release!

Purpose of this release is to show how FISTA is working.

Main features

• DataContainer class, responsible for executing all matrix/vector operations
• FISTA solver for Poisson noise
• Test suites to validate changes