closes#2244
- `mnorm` has been renamed to `normalize-matrix`
to reflect what it actually does, which
is normalize a matrix, not find a norm
of a matrix.
- `mnorm` is no longer a word defined here.
- bugfix: previously, `normalize-matrix` found
the supremum of a matrix (`mmax`),
before taking the supremum's absolute
value (`abs`) and dividing the matrix
by it (`m/n`).
for matrices containing only negative
values and 0, the supremum is 0, and
a `div-by-zero` error was thrown.
`normalize-matrix` has been fixed to
first `abs` all the matrix elements,
and then find the supremum and divide,
it also receieved a zero-matrix? guard
for optimization and preventing
`div-by-zero`.
- new alias: `hilbert-schmidt-norm` for
`frobenius-norm`, to go along with
`math.matrices.extras.<hilbert-matrix>`
and improve searchability by physicists.
- new word: `matrix-p-norm`, written as an
analogue of `math.vectors.p-norm`.
- new word: `matrix-p-q-norm`, which generalizes
entrywise matrix norm over the L^p,q
vector space.
- new word: `matrix-p-norm-entrywise`:
`matrix-p-norm`'s fallback
for p =/= 1, 2, inf; analogue of
`math.vectors.p-norm-default`.
- all norm words have gotten new docs,
`zero-matrix?` guards as an optimisation,
and most have gotten new tests.
math.matrices.elimination: move to extra
math.matrices.extras: expand with esoteric, less-used and unfinished code from basis
- math.matrices and .extras receive more words, tests, and docs
- matrix has become a predicate class
- 94% of matrices words have complete docs
- 77% of matrices.extras words have complete docs
- much more consistent naming for constructors etc
- added missing words / features such as main-diagonal and anti-transpose
- optimizations
- lots of documentation
This is for calculating e^x-1 for small values more accurately. You can also
call expm1(x) function if you want, and it's available on your platform.
FUNCTION: double expm1 ( double x )
IN: scratchpad [
{ byte-array } declare
[ 0 alien-unsigned-4 32 shift ]
[ 4 alien-unsigned-4 ] bi bitor
64 >signed
] optimized.
! working
[
dup >R 0 alien-unsigned-4 32 fixnum-shift
R> 4 alien-unsigned-4 over tag 0 eq?
[ fixnum-bitor ] [ fixnum>bignum bignum-bitor ] if
18446744073709551615 >R >bignum R> bignum-bitand
dup 63 bignum-bit? [ 18446744073709551616 bignum- ] [ ] if
]
! broken
[
dup >R 0 alien-unsigned-4 32 fixnum-shift
R> 4 alien-unsigned-4 over tag 0 eq?
[ fixnum-bitor ] [ fixnum>bignum bignum-bitor ] if
dup 63 bignum-bit? [ 18446744073709551616 bignum- ] [ ] if
]
The second case correctly eliminates the bitand but incorrectly assumes
that the item on the stack (which is an integer -- either a fixnum or a
bignum), was converted to a bignum.
Addresses #2170
- fixed: `interval-bitor` caused bit-growth
- improved: `interval-bitor` more exact about lower bounds
The added utility words could be used as a basis to make the other bitwise
interval operations more exact also.
Make both `empty-interval` and `full-interval` singletons, use generic functions
and methods where they are special-cased.
All words which work with interval points should also now work with the special
intervals.
Afaik, the mean word works both for population data and for
samples. Which is different from standard deviations, where you have
different formulas for population and sample std.
John Benediktsson
Doug Coleman
Cat Stevens
timor
Alexander Iljin
Björn Lindqvist