On Sun, Apr 26, 2026 at 03:14:15PM -0400, funkymail via cypherpunks wrote:
On Sun, Apr 26, 2026 at 03:10:06PM -0400, funkymail via cypherpunks wrote:
> > > > > > > > > https://ar.anyone.tech/6cb-OyzPFi1nJYZ-uHAHH2uwoCbosf0AQuS_qGEgBXY#-U_3AcjRg... > > > > > https://ar.anyone.tech/yYCz9PNR58vqF0oJI0bivwVmiOJ4Nn3Bcelast6ynVw#j1dSORCMP... > > > > https://ar.anyone.tech/Ktgtg2uyYl7bQqkEhApVkCGVZi0EpPRPMKU4QXEH4Oo#STcOlX11A...
> > > https://ar.anyone.tech/yz5qwtbaNxyukTrFyaO3YDFEcwweUyBvumiEDtvVnQg#x36BlkgBI...
> > gotta a big strong stopping associated with "trend quickly and > > reliably to" might relate to stopping fast AI > > very strong > > > > https://ar.anyone.tech/4JsidwtRdeBy8DCz0zj1jOPzjolbprJBTpcoGjLe7CQ#x36BlkgBI...
> > but i got more optimization and tech techniques from the stopping :) > > new stuff: > if you want to find a signal with a known kind of mixing, you can use > _all_ the metrics you have that identify the signal (for FastICA the > gaussianness is minimized to desum the signals, summed signals are more > gaussian) and mathematically look at how the numbers of that metric > directly relate to the mixing parameters (the derivative for example, > somewhat corrupt lost part) to "invert" it. > > hard to say straight > > but the positivity is helpfup for trying to continue, got a really > big stop though :s > > https://ar.anyone.tech/bEFskvOegFTcuLAV06d2baz1JE_BBNoInj2S5-_oncI#x36BlkgBI...
termux closed unsure where or if my draft is ... hmm they must be temporary editor files
i'm at https://web.archive.org/web/20260419233218/http://research.ics.aalto.fi/ica/... . first i went to wikipedia and found a paper introducing fastica and followed its footnote.
https://ar.anyone.tech/J9Kqw2qRW-1ALpz0-RHgJpoSe9kGlu33HDtpL9WOCFk#YphTM_TEq... https://ar.anyone.tech/cd_I0FskafQts2C0_jJC7XUldH5HMCpL-7xclGmH_uc
[fill-in: so, let's get a copy of fastica]
i ended up going to https://github.com/derekharrison/fastICA_C . it looks more familiar to me (which also means more triggering and challenging i'm afraid) and may be the actual code i started working off of (although the dates now don't seem to line up it could still work out as we had a lot of timestamp falsification going on) [but i think the one i had had more documentation could be wrong tho]
~/fastICA_C $ find * ReadingData.m inc inc/AlgoFunctions.h inc/Functions.h inc/MatrixOps.h inc/Memory.h inc/Parameter.h inc/ParameterDefinition.h lib lib/AlgorithmFunctions.c lib/ExportingData.c lib/Functions.c lib/MatrixOps.c lib/Memory.c lib/Setup.c src src/ICAMain.c src/ParameterInput.c src/Sources.c
let's draft a buildfile, a makefile. hrm this might take me a bit. patsubst? wildcard?
CFLAGS=-Iinc fastica: $(patsubst %.c,%.o,$(wildcard src/*.c)) libfastica.a libfastica.a: $(patsubst %.c,%.o,$(wildcard lib/*.c))
i may need more things specified too
https://ar.anyone.tech/w-qCfpCBqFg2_SKFP6E-o6rYU_JOCQxTZ9OUkb7gZcI#QwJoMKDMM... https://ar.anyone.tech/aGenbOwZwY0p2iVyY7RfRHTPPqCNlSXxjmEBbo-gbvA
here's where i'm at. i'll ask a business AI why my static library can't use the the c standard library.
$ cat makefile; make 2>&1 | head -n 16 CFLAGS=-Iinc fastica: $(patsubst %.c,%.o,$(wildcard src/*.c)) libfastica.a $(LD) $^ -o $@ libfastica.a: $(patsubst %.c,%.o,$(wildcard lib/*.c)) $(AR) cru $@ $^ %.o: %.c makefile clean: rm fastica libfastica.a */*.o cc -Iinc -c -o src/ICAMain.o src/ICAMain.c cc -Iinc -c -o src/ParameterInput.o src/ParameterInput.c cc -Iinc -c -o src/Sources.o src/Sources.c cc -Iinc -c -o lib/AlgorithmFunctions.o lib/AlgorithmFunctions.c cc -Iinc -c -o lib/ExportingData.o lib/ExportingData.c cc -Iinc -c -o lib/Functions.o lib/Functions.c cc -Iinc -c -o lib/MatrixOps.o lib/MatrixOps.c cc -Iinc -c -o lib/Memory.o lib/Memory.c cc -Iinc -c -o lib/Setup.o lib/Setup.c ar cru libfastica.a lib/AlgorithmFunctions.o lib/ExportingData.o lib/Functions.o lib/MatrixOps.o lib/Memory.o lib/Setup.o ld src/ICAMain.o src/ParameterInput.o src/Sources.o libfastica.a -o fastica ld: error: undefined symbol: time
> referenced by ICAMain.c > src/ICAMain.o:(main)
ld: error: undefined symbol: srand
https://ar.anyone.tech/0g1JiCRU1bwt3S_HceDdThNmmrcAAtMS-T6delVdsA0#QwJoMKDMM... https://ar.anyone.tech/ZPYTCDHgXfWUTBv0w0ajWyOi-Bv0gFQ6HCLZLc9Z7do
here's working makefile on termux:
CFLAGS=-Iinc fastica: $(patsubst %.c,%.o,$(wildcard src/*.c)) libfastica.a $(CC) -lm $^ -o $@ libfastica.a: $(patsubst %.c,%.o,$(wildcard lib/*.c)) $(AR) cru $@ $^ %.o: %.c makefile clean: rm fastica libfastica.a */*.o
when you run the resulting fastica binary it dumps numbers.
let's see its source.
oh hum
/* * ICAMain.c * * Created on: 14 apr. 2016 * Author: Derek W. Harrison * * This code is a C implementation of the FastICA method for recovering independent * components in component mixtures * * For more information refer to the paper 'ICA: Algorithms and applications' which is found * in the same directory as the source files of this code. */ #include "../inc/../inc/Parameter.h"
this isn't the repo this code came from, google gave me a mutated copy from a few years later. explains the strange dates some.
https://ar.anyone.tech/IxdnrOkxZceLrPsmIAFjIPjN_XMmqn5caa6HmPV06RA#PPPPuNFDc... https://ar.anyone.tech/4fhFx4Bq6oaxKeDIdl1Ob3G91vjcZtjoIshQg0MGbL8
ok so the main function offloads each step to a named function and it looks like those are declared in the inc folder with the strange relative path but they'd mostly be defined in src
/* * ParameterInput.c * * Created on: 14 apr. 2016 * Author: dharrison */
#include "../inc/ParameterDefinition.h"
void ParameterInput() { /* * The user parameters are entered here. This is the only section of the code * Which should be changed by the user. Note that 6 sources are available. If more sources * are required or desired these need to be entered in the corresponding function SetUpSources() * in Functions.c. The extra sources need to be added in Sources.c * If less are required or desired, the excess sources need to be commented or removed * from the SetUpSources() function. */
All the data is hardcoded, and doing so is spread across multiple files.
Since the goal is understanding the algorithm for using it, I have interest in just copying it over. But I'd also like to make sure it works, so maybe I'll stick with his test dat--
https://ar.anyone.tech/4eXjDBcP5QqMGwIMEn2Sccx2wn2bfgRVWFIAvaijHSU#2fgdvqysn... https://ar.anyone.tech/nnI2hEysA48eue5U0phRPPbB7YpC7EhwjA9n-ybXAlw
This sourcebase looks kinda messed up or corrupt to me, despite clearly functional, some of the user parameters are in the lib folder.
I'll visit the C++ implementation that's still linked on pages.
https://ar.anyone.tech/YS9ThIG9YUL_WwrRtN-p_mlTQCL8Xnzf-0tainEAhfw#2fgdvqysn... https://ar.anyone.tech/5DAOVj_KdGYmQuGehpVyUOUlvq1Qql_tDsL8QtXcxQ8 I am not affiliated with https://ar.anyone.tech .
I said this to google.com: i'd like a quick and simple standalone fastica implementation with source for learning, something that will build on termux. i searched for fastica c but got a bad republish with missing files and tangled layout. and its ai mode sent me to https://gist.github.com/Tetraquark/5bf052134d413f79748ce3aceaaf42df i have now installed matplotlib in termux with its dependencies. this may have used me up for this :s apologies. https://ar.anyone.tech/j4UPw7MyJS9z4QEUaGxGM-lnROB6RqY1F5wVPElUn5M#cYIoJyJe0... https://ar.anyone.tech/cryS0AT9Qx92qS-FirWlbJMtvPw3V1D8izKZch6D0TI I am not affiliated with https://ar.anyone.tech .