-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathCHANGES.txt
125 lines (124 loc) · 5.39 KB
/
CHANGES.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
v1.3.4, 10/02/2015
- Corrected direct setting of bandwidth / covariance
- Added doc for convolution
v1.3.3, 19/8/2014
- Correct use of kernel in non-parametric regression
- Added attributes to BootstrapResult
- Added efficient binning in Cythong and Python
- If using Cython, linear binning is used, which is better for small bin numbers
- Using rfft to save time (computes half the values)
v1.3.1, 8/8/2014
- Removed necessity of calling fit
- Added warning to main text
- Small FFT optimization
v1.3.0, 8/8/2014
- Removed unnecessary type conversion
- Rewrote the non-parametric fitting
- Rewrote parametric fitting
- The interface should be now uniform across the board, but incompatible with previous versions
v1.2.5, 6/8/2014
- Using 'out' rather than 'output' for in-place operations to conform to numpy convention
- Added a kernels.Kernel base-class implementing cdf, pm1, pm2, fft and dct in term of the pdf
- Corrected all the kernels, both Python and Cython
- Added and restructure kernel tests
- Restructured KDE tests.
- Moved TransformKDE to the kde_methods
- Updated tutorial
- Added decorators for 1 argument numpy transformations (i.e. functions working
on arrays and returning an array of same shape as the input)
v1.2.4, 30/7/2014
- Corrected KDE code to pass tests
- All tests pass for Python2 and Python3
- Corrected loader for python 3.4
v1.2.2, 30/7/2014
- Corrected distribution + version + CHANGES
v1.2.0, 30/7/2014
- Corrected in kernel_smoothing the use of atleast_?d to allow non-ndarray arguments
- Changed interface to select KDE methods to use objects
v1.1.17, 29/7/2013
- Corrected kde.KDE1D initialization and weigth handling
- Corrected CSV output of fitting
- Corrected the kde.botev_bandwidth to work with small input sizes
v1.1.16, 28/5/2013
- Remove need for cython in any module (was partial)
- PEP8 compliance
v1.1.15, 28/5/2013
- Changed description
- Doc now hosted at readthedocs.org
- Hosting package at PyPi
v1.1.14, 27/5/2013
- Unifying code base using a compatibility module
- sharedmem.py is now the same for Python2 and Python3 (and working with python3!)
- cleaned and corrected CSV output code for python 2 and python 3
v1.1.13, 21/5/2013
- Removed reference to npy_float128. Seems it doesn't work on 32 bits machines.
- Adapted to 32 bits machines
- Changed integration from basic to Simpson's method
- If cython is not present (on not wanted!), kernels exist also as pure python
functions
- Free switching between pure python and cython kernels
- Cython 1.19 is not really needed, so downgrading the need to 1.17
- Moved Cython dependencies into an "extra" with name 'Cython'
v1.1.12, 15/5/2013
- Added transformed KDE class, to compute KDE using variable change
v1.1.11, 27/2/2013
- Correct CDF of tricube kernel (error inducing a 1e-4 error on CDF computation)
- Added tests for kernels
- Added new interface for 1D local polynomial -- compatible with previous one!
- User can specify their own design matrix to allow for non-polynomial basis
v1.1.10, 26/2/2013
- Added Epanechnikov kernel
- plot_residual_tests now supports nD regression
- corrected kde in case no method is specified
- corrected grid_eval (now works)
- corrected weights application
v1.1.9, 18/2/2013
- Added tests for the various modules
- Corrected bug in sharedmem for Python 3
- Now using print_function and relative_import futures to minimise changes between Python 2 and 3
v1.1.8, 29/1/2013
- Corrected Cython import error in kernels.py
- Added "from __future__ import division" in all numerical module
v1.1.7, 28/1/2013
- Corrected installation bug
- Started tutorial on non-parametric fitting
- Corrected a few bugs in Spatial Averaging class
v1.1.6, 27/1/2013
- Corrected a few issues with documentation to allow for PDF generation
- Changed the __init__.py to remove references to pylab, so it works without GUI
- Corrected grid code to account for weights and lambdas
v1.1.5, 25/1/2013
- Added CDT and FFT-based KDE estimation
- Added TriCube kernel for demonstration
v1.1.4, 23/1/2013
- Refactored 1D KDE estimation to support multiple methods.
- Refactored documentation to reflect this
v1.1.3, 20/1/2013
- Added nD local polynomial regression
- Added 1D bounded KDE estimation
v1.1.2, 11/1/2013
- Added 1D local polynomial regression
v1.1.0, 8/1/2013
- Curve fitting changed to match non-parametric interface
- Unified bootstrap
v1.0.23, 6/1/2012
- Adapted the CurveFitting to fit the new bootstrap
v1.0.22, 6/1/2013
- Changed interface to add user functions, using objects now
- Added start of documentation
- Changed function and residual module loader to avoir name conflicts
- Corrected the loading of user-defined functions
v1.0.21, 5/1/2013
- Improved description
- Changed dependancy from setuptools to distribute because of a problem with setuptools 0.6
v1.0.20, 5/1/2013
- Made non-parametric bootstrap method parrallel
- LocalLinear1D is done in Cython now
- Use LocalLinear1D for residual analyses rather than SpatialAverage
- Changed bootstrap method to be more general.
- Changed the functions and residuals modules to read from the PYQTFIT_PATH environment variable the list of extra folders to analyse
- Added Local Linar Kernel Regression method
v1.0.19, 10/12/2012 - Corrected dependency specifications.
v1.0.14, 4/12/2012 - Added version numbers for the requirements
v1.0.10, 4/12/2012 - First release distributed with pip
v1.0.9, 4/12/2012 - Last release without pip