| View previous topic :: View next topic |
| Author |
Message |
erach27@gmail.com
Guest
|
 Posted: Sat Mar 22, 2008 4:09 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
 |
|
On Mar 21, 4:12 pm, David <davidd...@gmail.com> wrote:
| Quote: | On Mar 21, 5:45 am, the_dawggie <the_dawg...@hotmail.com> wrote:
On Mar 21, 9:33 pm, "erac...@gmail.com" <erac...@gmail.com> wrote:
hi,
i am now trying for 3 days rubbing old urine on body in the bathroom,
keeping it on for 15 minutes and then washing it out with cold water,
and drinking new urine.
ramniklal gada (urine therapist, mumbai 9820600741) has told me that
if u put old urine in the hair it stands up. so i guess i should use
diluted old urine (not fresh urine --- it turns the hair yellow).
and washing face and eyes closed at that time with old urine.
yesterday, i slept well even though my meds slightly reduced. i am
reducing meds most gradually. over months.
Your shows will smell.
fathom that
|
Hi, I was going in mania. But nobody smelt my old urine on the body.
I guess I should apply some of it on my body and then wash it off.
Now, I hope I get my typical anti-psychotic (Trinicalm Plus) because
it is not available at my regular chemist. My tranquilizer + sleeping
pill side-effect Chlorpromazine (Zicot 25 mg) is not available on the
market.
Erach
Erach |
|
| |
|
 Back to top |
%
Guest
|
| Posted: Sat Mar 22, 2008 9:17 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
| no |
|
| |
|
| Back to top |
Guest
|
| Posted: Tue Mar 25, 2008 5:30 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
Try cun in your hair. It works better.
http://startelegram.typepad.com/sky_talk/2008/03/woman-files-law.html |
|
| |
|
| Back to top |
Guest
|
| Posted: Tue Mar 25, 2008 5:35 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
I can't believe I was looking for AI info when I ran into this. Well,
it reminded of the news story I provided the link to. If that woman
was a hot Latin babe and she fell asleep on an airliner, she was
asking for it, if you ask me! You know, driving while black means you
should get arrested and sleeping in a public place while being a hot
latin chick means you should get what's cumming to you. Being an
intern in the Clinton White Stuff House means....well, I could go on
and on.... |
|
| |
|
| Back to top |
erach27@gmail.com
Guest
|
| Posted: Tue Mar 25, 2008 5:37 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
On Mar 25, 10:35 am, giveitawhril2...@gmail.com wrote:
| Quote: | I can't believe I was looking for AI info when I ran into this. Well,
it reminded of the news story I provided the link to. If that woman
was a hot Latin babe and she fell asleep on an airliner, she was
asking for it, if you ask me! You know, driving while black means you
should get arrested and sleeping in a public place while being a hot
latin chick means you should get what's cumming to you. Being an
intern in the Clinton White Stuff House means....well, I could go on
and on....
|
heard of "urine therapy" 156,000 web pages on google.com
which cures all these diseases or so they claim.
check http://spacetimemotion.tripod.com too
trust your doctor, going to visit my psychiatrist on saturday,
i am manic-depressive bipolar dysphoric.
erc |
|
| |
|
| Back to top |
Guest
|
| Posted: Tue Mar 25, 2008 5:37 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
| "Try cun in your hair. " I meant "cum" |
|
| |
|
| Back to top |
%
Guest
|
| Posted: Tue Mar 25, 2008 10:39 am Post subject: Re: (urine therapy 150,000 web pages) rubbing old urine on b |
|
|
giveitawhril2008@gmail.com wrote:
| Quote: | "Try cun in your hair. " I meant "cum"
|
its not working either way ,
your delievery is totally shot |
|
| |
|
| Back to top |
Santeri Pettersson
Guest
|
| Posted: Wed Jun 11, 2008 7:14 am Post subject: Ryhmän sfnet.atk.viritys virallinen kuvaus |
|
|
YLE radion aamu-uutisissa 02.06.2008 kauhisteltiin taasen maatamme jo toista
vuotta kohdanneen totaalisen mehiläiskadon megalomaanisuutta.
Poissaolollaan tähdensi viestiä aiemmat bakteeri-, virus-, varroapunkki- ja
ties mitkä kiinalaisapuruokamyrkytysdisinfot! Tässä on varsin tärkeää
huomata, että vaikka asiaa on kansainvälisin paniikitasoin tutkailtu ja
laboraatioissa etsiskelty syyllistä massoittain, ei vaan aukea syyt! Tai no
OIKEASTI tieto on esitetty jo vuosia sitten perustuvaksi
ydinvoimalapäästöjen säteilyionisaatiomekanismeihin. Mutta koskei ydinala
halua vastata rikoksistaan yhä vaan faktaa pantataan, paitsi toki lähinnä
meikä täällä!)
Yhtä kaikki siis tuhoa on kertynyt -90% katoon maapallolla ja KAIKKIALLA!
Ydinyhtiöt yrittivät valtavalla paineella kyltymättä valehdella, mutta koska
tieto säteilöytaustasta pääsi leviämään ei homma ottanut ydinalan SUUREKSI
harmiksi enää tuulta. Suuri osa maailman väestä siis tietää, ettei yhä
hengissä olevat ja sielä täällä pörräävät mehiläiset kuole nähdysti
tauteihin. Koskei sairaita ole vuosiin kyetty edes näyttämään! Ainoaksi
järkeväksi selitykseksi on vääjäämättä jäänyt terveitten mehiläisten
hahmotetut lentoöksymät säteilyioniosaation prässääminä. On jännä tosiaan
nähdä. Koska ensimmäiset lehdet tuovat ydinfaktaa pöytään maailmalta?
PS. Kuvaavaa tilanteessa on, että ilmatieteenlaitoksen salaamat
ionosfäärianomaliakarttatiedot ovat mitä tarkimmin poistettu yleisön
käytöstä. Eivät "roistouksissaan" edes vastaa asiallisiin kyselyihin
aiheesta!! |
|
| |
|
| Back to top |
Kicn
Guest
|
| Posted: Mon Jun 16, 2008 3:10 pm Post subject: Re: [Fwd: Re: Drawing NN diagram] |
|
|
Sorry, wrong forwarded email.
Kicn wrote:
| Quote: |
------------------------------------------------------------------------
Subject:
Re: Drawing NN diagram
From:
Kicn <Kicn2008@gmail.com
Date:
Mon, 16 Jun 2008 18:08:07 +0800
To:
steve@tropheus.demon.co.uk
To:
steve@tropheus.demon.co.uk
Newsgroups:
comp.ai.neural-nets
Path:
news.cuhk.edu.hk!not-for-mail
Newsgroups:
comp.ai.neural-nets
Organization:
The Chinese University of Hong Kong
Message-ID:
48563B87.4010001@gmail.com
References:
g350q7$rj1$1@ijustice.itsc.cuhk.edu.hk
psac54hroiel6me5fdv5n218424n011ggi@4ax.com
NNTP-Posting-Host:
pc135.mae.cuhk.edu.hk
MIME-Version:
1.0
Content-Type:
multipart/mixed; boundary="------------020504040502070100040507"
X-Trace:
ijustice.itsc.cuhk.edu.hk 1213610641 1575 137.189.101.135 (16 Jun 2008
10:04:01 GMT)
X-Complaints-To:
abuse@cuhk.edu.hk.removeme
NNTP-Posting-Date:
Mon, 16 Jun 2008 10:04:01 +0000 (UTC)
User-Agent:
Thunderbird 2.0.0.14 (Windows/20080421)
In-Reply-To:
psac54hroiel6me5fdv5n218424n011ggi@4ax.com
Xref:
news.cuhk.edu.hk comp.ai.neural-nets:76393
Thanks all. I took a look on those softwares but it seems that they
can't draw diagram like the one I attached. (that is adapted from M.T.
Hagan's Intro to NN)... any advice please?
Stephen Wolstenholme wrote:
On Mon, 16 Jun 2008 14:26:04 +0800, Kicn <Kicn2008@gmail.com> wrote:
Hi all, is there any tools to draw NN diagram, something like
MathType to write equations? I am currently using MS Word's graphic
tool to draw but found it not very convenient.. some advice please?
EasyNN-plus has a few printable displays including a network diagram.
Steve
------------------------------------------------------------------------
|
|
|
| |
|
| Back to top |
Lassi Hartmann
Guest
|
| Posted: Mon Jul 14, 2008 2:36 pm Post subject: Re: Report: Ahmadinejad Tells Japan to Prepare for a World W |
|
|
01.05.2008 Jää historian kirjoihin erityispäivämääränä josta alkaa ydinalan
vaikeudet kasautumaan aivan erityisesti myös Suomessa. Jo siitä asti kun
puollisentoista vuotta sitten Krister Fuuglesangin NASA:n
avaruusionisaatioraportin tiedot maapalloa kohtaavasta vääjäämättömästä
ydinvoimapäästöjen megatuhosta vuodettiin maailman päättäjille ydinalan
alasajon välttämättömyys on ollut tiedossa. Samasta pohjasta jopa IPCC
ennakoi koko ilmastojärjestelmän tuhoa 10v sisällä. Mutta toukokuusta alkaa
myös tietomuruja mainiyusta maailmanhistorian suurimmasta
ydinkatastrofitulemisesta vuotaa, esim. kauttani viimein suuren kansanmassan
tietoisuuteen. Mistä tarkemmin on kyse. Maailman metastabiilinen
perustaustataso on 10-kertaistunut totaalisen tuhoisaan suuntaansa
romahdellen. Luonto itsessään viestittää katastrofaalisista
mehiläistuhoviesteistä ja vastaavantyyppisistä
ydinkampamaneetti-invaasioista kiihtyvästi. Einsteinin 5v
mehiläishäviämisnäytöistä ennakoima ionosfäärituho on vyöryämässä kaikin
merkein ihmiskunnan kohtaloksi.
Onnistuin saamaan ydinalan silkaksi kauhuksi YVA 4/TEM tiedostoihin
maininnat siitä, että tulevat sukupolvet voivat myös arkistofaktojen
perusteella tuomita nykyiset ydintoimitsijat, firmat ja päättäjät ydinalan
kansainväliseen Nyyrbergin oikeudenkäyntiin päättäjiemme nyttemmin kiistatta
tietämistä rikoksistaan ihmiskuntaa vastaan. Niin pitkälle on prosessi
edennyt, että Sisäministeriö jo panikoi tämän tulevan faktan tietäen!
YLE-teksti-TV kirjoitti, miten vähin äänin oli muutettu maamme lakia
rajusti: " Rikoksista ihmisyyttä vastaan on toimeenpantu lakimuutos 01.05
lukien, että jatkossa henkilöitä, ryhmiä ja teollisuuslaitoksia jne. jotka
tuomitaan erittäin raskauttavista (ydin) ja sotatoimista kansaansa vastaan
ja rikoksista ihmisyyttä vastaan ei voida tuomita Suomesta kansainvälisiin
Haagin tuomioistuimiin saamaan jopa kuolemantuomioitaan. Vaan jopa näin
rajuista rikoksista langetetaan korkeintaan Suomen lainsäädännöllä paikan
päällä maksimirangaistukseksi rajaten vain elinkautisiksi! Siis huh, hu
alkaa ydinherrojemme polvet tutista alati kasvavien ydinrikosilmitulojen
paineissa! Ilmankos Heiniluoma sanoi, ettei Suomeen TVO:n kaavailemia
isotooppilaitostartteja ja ydinvoimaloittemme keskeisiä
plutoniumpoltto-oikeuksia, uraanikaivosmassiiveineen ja
ydinjätekäsittelyineen EU:sta kiirehditä. Vaan oltiin yht äkkiä sitä mieltä,
ettei edes seuraavakaan hallitus tunke kokoomuksen/TVO/Posiva:n
suunnattomaksi mieliharmiksi lupaa hakien kansantahtoa vastaan
edellämainittuihin NATOn hyväksymiin-optiotoimiin.
02.04-08. Uutiset kertoivat, että jälleen kerran oli Norjassa 200km Oslosta
pohjoiseen Ottessa sortunut tektoniliikunnasta 300m pitkä ja 50m leveä
kalliorinne ja allle oli hautautunut vähintään 150 hengen pakolaisvirran
startanneen väestön kymmenet talot. Edellisestä tästä kuuluisasta 2500 km
pitkästä Laatokka-Skotlantitektonisesta Litoraanisaumalikunnasta Olkiluodon
läpi olimme saaneet viestejä Norjasta 5 hengen kuolemaepäilystä ja 7m
liikuneesta vuorirojahduksesta kuukausi sitten. Päivän selvästi tämä
kyseinen maailman syvin maanpäällinen tektonisauma on hurjasti
aktivoitumassa. Niin ruotsalaisvoimalat, kuin Olkiluodon kaikki laittomasti
mm. kallioon ankkuroidut ydinvoimalat tulevat liikuntojen murskaamina
joutumaan suunnattomiin vaikeuksiin. 5cm/v liikuva sauma kun tuottaa
jatkossa ruotsalaisopein huikeita mitattuja yli 8 Rihterin maanjäristyksiä!
Edes vuoden takaista 450m kallioliikuntoa Ruotsissa saati saman aikaista 14
Norjan luolan tuhoa ei olla uskallettu tutkia julkisesti. Virannoimaisten
panikointi muutti jo nyt koko Ruotsin tektonisaumaan ydinjätehautauskuviot
totaalisesti syväporausmalleihinsa. Esimauksi myös Suomeen 10 kertaistuneine
lisämaksuin turvarajamuutospakoin. Mutta kaikesta näkee, ettei luonto tähän
tyydy, vaan pahentaa odotettaan alati! |
|
| |
|
| Back to top |
Greg Heath
Guest
|
| Posted: Mon Aug 11, 2008 5:32 pm Post subject: Re: Regression with too large training set. does it lead to |
|
|
On Aug 11, 12:30 pm, Greg Heath <he...@alumni.brown.edu> wrote:
| Quote: | On Aug 11, 12:33 am, VijaKhara <VijaKh...@gmail.com> wrote:
Hi all,
I am using linear regression to fit the data to a straight line. I
have two observed sets T1 and T2, T1 is larger than T2.
If I use T1 as the training set and T2 as the testing set, I get a
larger mean testing error than if I use T2 as training set and T1 as
the testing set. Also the latter provides me the better testing error
if I test with all data set (T1+T2). I am wondering what is the reason
for this. Is it true that training the model with too large training
set may lead to overfitting?
No. Just the opposite. The larger the training set, the more
accurate the weight estimates. The general rule of thumb is
Ntrn >> p+1. Practical ratios of N/(p+1) are typically larger
than 10. Estimation of weight estimate uncertainty involves
the inversion of the data matrix. You will have to look up the
details.
On the other hand, given the weights, the larger the test set,
the more precise the error estimates. If you assume errors
are Gaussian distributed, MSE is Chi-SQ and you can
estimate Ntst for a given precision.
So, the partition of N = Ntrn + Ntst involves a tradeoff,
which can be easily satisfied if N is large enough. If it
is not large enough, resampling techniques (e.g.,
10-fold XVAL or bootstrapping) have to be used.
Unfortunately, you have neglected to provide important details:
Number of predictors, p
Number of responses, (I assume 1 ?)
Size of T1
Size of T2
Desired precision
If you feel that T1 and T2 don't seem to represent random
draws from the same probability distribution, you may want
to combine them and then randomly pick new training and
test sets. Averaging over multiple trials should give you more
confidence w.r.t. estimates of mean+/-stdv.
10-fold XVAL or bootstrapping are other options.
|
Please do not send separate copies of the same post to
multiple groups. Instead, send one copy and crosspost by
putting multiple group names in the "TO" or "NEWSGROUP"
line.
I have sent this reply to two other groups by crossposting.
Hope this helps.
Greg |
|
| |
|
| Back to top |
Greg Heath
Guest
|
| Posted: Thu Aug 14, 2008 2:46 pm Post subject: Re: About train Neural Network function |
|
|
On Aug 13, 12:57 pm, "Manos " <ezoul...@in.gr> wrote:
| Quote: | Hi my friends,
I use neural nets and I try to use the function train.
I know function has the following syntax
[net,tr,Y,E,Pf,Af] = train(net,P,T,Pi,Ai,VV,TV).
I could not found the meaning of VV and TV parameters.
|
VV and TV are structures that are created by the assignments
VV.P = valin;
VV.T = valout;
TV.P = tstin;
TV.T = tstout;
| Quote: | Is it validation and taining part?
|
No. Validation and testing.
| Quote: | What should be the
dimensions of P,T,VV,TV.
|
N = Ntrn + Ntst + Nval
Ntrn must be large enough to obtain accurate weight
estimates.
Ntst must be large enough to obtain precise error
estimates.
Nval must be large enough to obtain error estimates
that are sufficiently accurate to sense when an
overfit (i.e., overparameterized) net is starting to
be overtrained.
Typically, for an I-H-O topology with Nw weights
(Nw = (I+1)*H+(H+1)*O), the ratio r = Ntrn/Nw is
chosen in the range ~2 <= r <= ~30 . Usually H is
chosen by trial and error starting with r ~10.
However, when VV is used to stop training, smaller
values of r can be used.
| Quote: | Is P for the training and T the
targets of training set only and VV for validation and TV
for test?
|
Yes.
| Quote: | Can you give me a meaning?
|
VV is used to stop training.
TV is used to obtain an unbiased estimate of generalization error.
| Quote: | What I would like to do is to give training, validation and
test sets to train function, as I seperated by a self made
function and not to use the inherent divirand or any other
ready from Matlab.
|
I have many posts in c.s-s.m and comp.ai.neural-nets regarding
partitioning and/or splitting data sets into trn/val/tst sets.
Search the archives in Google Groups.
Hope this helps.
Greg
P.S. Crossposted to comp.ai.neural-nets |
|
| |
|
| Back to top |
Greg Heath
Guest
|
| Posted: Sat Aug 16, 2008 4:10 am Post subject: Re: help with testing and validating data |
|
|
On Aug 15, 7:04 pm, chan <chantelojinn...@yahoo.co.uk> wrote:
| Quote: | Hi i'm new to matlab and I need help with testing and validating my
data. I have split my data
50/50 and have already used one half as training data. I have trained
the data with the following code but i dont know how to test and
validate it in matlab. Can anyone please help me?
NN=network;
NN.numInputs=1;
NN.inputs{1}.size=2;
NN.numLayers=2;
NN.layers{1}.size=2;
|
The numbers of hidden nodes, H, may need to be
determined by trial and error.
| Quote: | NN.layers{2}.size=1;
NN.inputConnect(1)=1;
NN.layerConnect(2, 1)=1;
NN.outputConnect(2)=1;
NN.targetConnect(2)=1;
NN.biasConnect(1)=1;
NN.biasConnect(2)=1;
NN.layers{1}.transferFcn='logsig';
NN.layers{2}.transferFcn='logsig';
|
Better to use 'tansig' and 'purelin' with
zero mean inputs.
| Quote: | NN.initFcn='initlay';
NN.layers{1}.initFcn='initwb';
NN.inputWeights{1,1}.initFcn='rands';
NN.biases{1}.initFcn='rands';
NN.layers{2}.initFcn='initwb';
NN.layerWeights{2,1}.initFcn='rands';
NN.biases{2}.initFcn='rands';
NN=init(NN);
|
The above code is unneccessary. See below.
| Quote: | load dat1.txt;
A=dat1;
I=A(:,1:2)';
T=A(:,3:3)';
NNA=sim(NN,I)
NN.IW{1,1}
NN.trainFcn='trainb';
NN.inputWeights{1, 1}.learnFcn='learngd';
NN.biases{1}.learnFcn='learngd';
NN.layerWeights{2, 1}.learnFcn='learngd';
NN.biases{2}.learnFcn='learngd';
NN.performFcn='mse';
NN.trainParam.lr=1;
NN.trainParam.goal=0.001;
NN.trainParam.epochs=10000;
%NN.trainParam.show=100;
|
It is much easier to use defaults first. If any are insufficent,
then just change them.
| Quote: | [NN, tr]=train(NN, I, T);
NNA=sim(NN,I)
NN.IW{1,1}
|
For advice and examples, go to Google Groups and search with
greg-heath pretraining-advice newbies
greg-heath close-all newff
%%% untested code %%%
close all, clear all, clc, k=0
load dat1.txt;
A = dat1';
P = A(1:2, ;
T = A(3,;
k=k+1,figure(k)
subplot(2,1,1)
plot(P(1,,T)
subplot(2,1,2)
plot(P(2,,T)
% You may want to standardize the rows to zero mean and
% unity standard deviation. Then replot.
doc newff
help newff
% H is chosen by trial and error
% Search with greg-heath Neq Nw
if H==0,
net = newff(minmax(P), 1,{ 'purelin'});
else
net = newff(minmax(P),[H 1],{'tansig' 'purelin'});
end
net.trainParam.goal = var(T)/100;
net.trainParam.show = 10;
net = train(net,P,T);
ytrn = sim(net,P);
etrn = T-ytrn;
MSEtrn = mse(etrn)
ytst = sim(net,Ptst);
etst = T-ytst;
MSEtst = mse(etst)
The 1:1 trn/tst split may be insufficient. Try m:1
for m > 1.
Hope this helps.
Greg
Please do not post multiple copies to different newsgroups.
Crosspost one copy to multiple newsgroups.
Crossposted to c.a.n-n and c.s-s.n |
|
| |
|
| Back to top |
chan
Guest
|
| Posted: Sat Aug 16, 2008 9:04 am Post subject: Re: help with testing and validating data |
|
|
Thanks for your reply. Im new to the groups and do not know how to
crosspost to multiple groups. |
|
| |
|
| Back to top |
chan
Guest
|
| Posted: Sat Aug 16, 2008 10:32 am Post subject: Re: help with testing and validating data |
|
|
I have tried the above code and it doesn't work. I need to train, test
and validate data using tansig and logsig. I can train the data from
the source dat1.txt using tansig and logsig using the code from my
first post, I just dont know how to test and validate the data. I am
completey new to matlab and really need help. |
|
| |
|
| Back to top |
|
