Some, but not all, tensors are representable as matrices. No, Reinassance Technologies [sp] uses a black box.
----- Original Message ----- From: "loveyourenemynow" <[EMAIL PROTECTED]> To: <amibroker@yahoogroups.com> Sent: Thursday, January 10, 2008 11:01 AM Subject: [amibroker] Re: extra info for each symbol for each bar, addcomposite on different timeframe > tensors are basically matrices > have you ever heard about Reinassance Technologies? Do u think they > use TA... ? > > > Thanks > Ly > --- In amibroker@yahoogroups.com, <[EMAIL PROTECTED]> wrote: >> >> Right on. >> Plus, of course, dynamic tensors. >> I'm sure it's more interesting, if less profitable, than trading. >> >> ----- Original Message ----- >> From: "loveyourenemynow" <[EMAIL PROTECTED]> >> To: <amibroker@yahoogroups.com> >> Sent: Thursday, January 10, 2008 12:54 AM >> Subject: [amibroker] Re: extra info for each symbol for each bar, >> addcomposite on different timeframe >> >> >> > Hi Mike, >> > thank you for your link. >> > I will try to look at it more carefully later, but just as a general >> > solution of the problem a function could be created to simulate >> > matrices of dimension 6(OHLCVI) X BarNumber (is new symbols can be >> > created from OLE) >> > By naming appropriately the new symbol, it can be related to the old >> > one via naming convention, and in principle also multidimensional >> > array could be handled by dynamically naming the symbols corresponding >> > to the dimensions (which is a way to implement matrices in AFL as >> > Tomasz pointed out somewhere) >> > Or may be the function could just work in terms of dynamically >> > variables for ex: >> > a[2,2] -? a_2_2 >> > so tha for example >> > >> > set(x,i,j,value){ varset(x+"_"+i+"_"+j,value);} >> > get(x,i,j){ return varget(x+"_"+i+"_"+j);} >> > (this is a just an ex, the syntax could be woring, but the idea is > there) >> > >> > so a[2,2]=5 would be set(a,2,2,5); >> > and a[2,2] get(a,2,2) >> > >> > using the symbol for a u immediately get the required extrad field per >> > symbol provided you loop through all the bars and set(sym,i) >> > >> > setA would be for vectors, setM for matrices >> > setM3 for tridimensional ararys and so .. >> > >> > i would be nice to also write some other functions like vectorial or >> > matrices opeerations >> > >> > I am sure somebody has already done that ... >> > >> > functions to get rows and columns could also be easly coded. >> > if operators could be overloaded it would be nice to code a[2,2] >> > directly .. >> > >> > >> > I will look into your code and thank you again >> > >> > Ly >> > >> > >> > >> > >> > --- In amibroker@yahoogroups.com, "Mike" <sfclimbers@> wrote: >> >> >> >> Ly, >> >> >> >> In case this is still regarding your post in message #118399 (how to >> >> trade only top percentile items under multiple criteria), I've gone >> >> and added a possible solution to that one for you. >> >> >> >> No, it's not a generic solution to extra data per symbol. But, maybe >> >> it will help for your percentile quest :) >> >> >> >> Mike >> >> >> >> --- In amibroker@yahoogroups.com, "loveyourenemynow" >> >> <loveyourenemynow@> wrote: >> >> > >> >> > For convenience I would like to store some extra info for each >> >> symbol, >> >> > and I would like it to be a vector, having a value for each bar, >> >> which >> >> > could be accessed at any time via foreign functions. >> >> > At the moment this is possible only for OHLCV ,and I think those >> >> > variable are protected). >> >> > I could create an extra ticker for each symbol,in which to > store up >> >> > to 5 info(using OCHLV), but I was thinking if there is a better > way. >> >> > >> >> > Basically it is equivalent to set up a table, so it can be achieved >> >> by >> >> > using an external plugin or object, but even in that case I would >> >> need >> >> > to store at least one info(a handle) to link to the correct >> >> row/column >> >> > in the external table, so at least one extra info is necessary any >> >> > way(unless the symbol is used but that you need a system to manage >> >> not >> >> > numerical indices for arrays, like a string list of this type >> >> > sym1,1,sym2,2 to scan through to get the right index) >> >> > >> >> > All this is rather cumbersome but conceptually is simply the use of >> >> > matrices. >> >> > >> >> > This is a very general problem, which if solved allows to do >> >> > basically to do anything we want with vectors and matrices, and it >> >> is >> >> > also a major limitation. >> >> > Also a related question: when addcomposite is used on different > time >> >> > frames does it have to be "scanned" each time in advance. >> >> > I thought time compression would automatically take care of it if >> >> > calculated just once at the lowest time resolution but I doens't >> >> seem so. >> >> > >> >> > Thanks >> >> > Ly >> >> > >> >> >> > >> > >> > >> > >> > Please note that this group is for discussion between users only. >> > >> > To get support from AmiBroker please send an e-mail directly to >> > SUPPORT {at} amibroker.com >> > >> > For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG: >> > http://www.amibroker.com/devlog/ >> > >> > For other support material please check also: >> > http://www.amibroker.com/support.html >> > >> > Yahoo! Groups Links >> > >> > >> > >> > > > > > Please note that this group is for discussion between users only. > > To get support from AmiBroker please send an e-mail directly to > SUPPORT {at} amibroker.com > > For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG: > http://www.amibroker.com/devlog/ > > For other support material please check also: > http://www.amibroker.com/support.html > > Yahoo! Groups Links > > >