-
Notifications
You must be signed in to change notification settings - Fork 18
/
revision_history.txt
86 lines (60 loc) · 3.06 KB
/
revision_history.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
ad-1.1.1
========
- Added ``gh`` utility function for generating gradient and hessian functions
of mathematical functions. The intent is to make it easy for users wanting to
use the ``ad`` package with``scipy.optimize`` functions.
ad-1.1
======
- Changed the core usage to be compatible with all base numeric types (int,
float, long, and complex).
- Changed the main constructor name from ``adfloat`` to ``adnumber`` because of
the full numeric support (not just for floats). The reasoning is because this
package should simply provide a "wrapper"-ish functionality for derivative
tracking (i.e., it shouldn't force the user to choose float objects if they
don't want to.
- Added ``real`` and ``imag`` properties to ``ADF`` class.
- Updated all admath functions to work with their respective numeric types using
the math and cmath module functions.
- Updated all admath function documentation to reflect basically the same as the
math module.
- Added ``pi`` and ``e`` module members to admath.
- Added ``atan2``, ``isinf`` and ``isnan`` methods to admath.
- Updated ``log10``, ``log1p``, ``expm1``, and ``fabs`` in admath to use the
actual math or cmath module functions, mostly for their higher precision at
certain input values (see the math/cmath module documentation for more
details).
- Bug fix: ``log`` didn't default to base ``e`` before. It's fixed now.
ad-1.0.3
========
- Fixed an issue with using Numpy arrays as input to ``adfloat`` in Python 3.3.
- Renamed ``_calculate_derivatives`` method to more appropriately be called
``_apply_chain_rule``.
ad-1.0.2
========
- Default constructor ``AD`` renamed to ``adfloat`` to follow general coding
guidelines better.
- Changed ``ADF(...)`` and ``ADV(...)`` to just ``ad(...)`` when printing out
the variables in either str or repr form (made a general printer--credits to
EOL!).
- Updated the ``_get_variables`` method to not depend on the ``d(...)`` method,
but to access the hidden member ``_lc`` directly for getting first derivative
variables.
- Renamed hidden member ``_hash`` to ``_trace``.
- Made the constructor for arrays more robust and now the tag is applied to all
objects that get created.
- Constructor is updated to return a copy of another AD object if supplied as an
input (i.e., ``x = adfloat(2); y = adfloat(x)``). This makes the two objects
equivalent, copying the nominal value and the derivatives, but naturally they
two objects are not identical (i.e., ``x==y`` returns ``True``, but ``x is y``
returns ``False``).
- Added ``sqrt`` method to ``ADF`` class because some ``numpy`` functions look
for this method of the objects it operates on (i.e., ``numpy.sqrt``,
``numpy.std``, etc.). Also updated the ``sqrt`` function in the ``admath``
module to actually use the ``math.sqrt`` method to perform the calculations.
ad-1.0.1
========
- Fixed some vectorization bugs
ad-1.0
======
- 1.0: Initial release. Nearly full differentiation support for all
math module functions.