Intervju sa Zdravkom Šurlanom, glavnim tajnikom Internacionalne lige humanista (Motrišta 45-6, 2009.)

 INTERVJU S POVODOM: prof. Zdravko ŠURLAN, glavni tajnik
Internacionalne lige humanista – lige za mir i svjetsko jedinstvo
Rat i tranzicija stavili su nas pred
velike izazove i nedoumice
MOTRIŠTA: Prof. Šurlan, ove godine navršava se 35. obljetnica Dubrovačke deklaracije o miru i povjerenju koju su, krajem studenoga 1974., napisali utemeljitelji Internacionalne lige humanista. Možete li nam, u kratkim crtama, reći - što je ILH danas i kako ćete obilježiti ovu značajnu obljetnicu?

 

ŠURLAN: Teško je u nekoliko rečenica sažeti sve ono što je prošla Internacionalna liga humanista - od ideje do danas. Ipak ću, u kratkim crtama, iznijeti samo neke od podataka, kako bi čitatelji mogli steći što cjelovitiju sliku o ILH.Ideja o pokretanju osnivanja novoga pokreta za mir i svjetsko jedinstvo – današnje Internacionalne lige humanista - potekla je od našega, u svijetu poznatog znanstvenika, filozofa, romanopisca, dramatičara (humanista 20. stoljeća) svjetskoga ugleda, blagopreminulog akademika Ivana Supeka (Hrvatska). Po rođenju humanist, po opredjeljenju antifašist, po djelovanju demokrat, nakon Drugoga svjetskog rata pridružio se međunarodnim pokretima za mir i dokinuće nuklearnog oružja. Sudjelovao je na mnogim međunarodnim konferencijama za mir s kojih je uputio mnoge apele, uvijek podržavajući načela općega i potpunog razoružanja. U cilju zalaganja za politički pluralizam (demokraciju) kao uvjet za razoružanje, mir i blagostanje, 1966. pokrenuo je časopis
Encyclopaedia Moderna kao glasilo za filozofiju znanosti i mir, zajedno s Jugoslavenskim pagvaškim društvom. Encyclopaedia Moderna imala je značajan utjecaj u cijeloj zemlji. I ne samo to, kao rektor Sveučilišta u Zagrebu od 1968. do 1970. i od 1970. do 1972. godine, poticao je reformu studija u skladu sa znanstvenom revolucijom i načelima humanizma. Na konferenciji Međunarodnoga udruženja sveučilišta u Montrealu, u ljeto 1970., zalagao se za veću suradnju u promicanju razumijevanja između raznih religija (različitih političkih sustava), što bi proširilo put prema svjetskom miru. Prihvaćen je
njegov prijedlog da se ustanovi mjesto za suradnju između sveučilišta i ubrzo je otvoren  Interuniverzitetski centar u Dubrovniku (1971) s više od 220 sveučilišta kao članova, kao primjer univerzalnoga duha znanosti i umjetnosti.

 
MOTRIŠTA: Dakle, Dubrovnik je mjesto rođenja Internacionalne lige humanista...
 

ŠURLAN: Točno tako. U Dubrovniku krajem studenoga 1974. godine, na inicijativu Ivana Supeka, sastao se bard mirotvoraca (Prvi međunarodni komitet humanista – univerzalista): Ava i Linus Pauling (dvostruki dobitnik Nobelove nagrade 1952. i 1962), Philip Noel-Baker (nobelovac), Aurelio Peccei (predsjednik Rimskog kluba), Sophia Wadia (književnica). Nakon deset dana intenzivnog rada u Interuniverzitetskom centru dali su analizu tadašnjeg svijeta i viziju željene budućnosti. Supekova načela humanizma sažeta u 10 točaka, kao sastavni dio Deklaracije, postali su polazište za rad i djelovanje Međunarodnoga pokreta za mir i svjetsko jedinstvo - Internacionalne lige humanista. Njima su dodana razmišljanja Abdus Salama (nobelovca), Dalai Lame (nobelovca), Guylord P. Hruwela (poznatoga istraživača iz Pensilvanije), koji, iz opravdanih razloga, nisu mogli nazočiti u razgovorima u Dubrovniku. Dubrovačka deklaracija razaslana je diljem svijeta i postigla je izvanredne pozitivne reakcije mnogih uglednih ljudi diljem svijeta, čiji su komentari pretočeni na Kongresu Svjetskoga
jedinstva u Filadelfiji.

 
MOTRIŠTA: Kad ste spomenuli Filadelfiju, što se sve događalo na Prvome svjetskom kongresu humanista?

 

ŠURLAN: „Biti ili ne biti, za ljude postojati, ovisi o našim sposobnostima da razvijamo sveopću civilizaciju koja će težiti miru, opstanku i humanim osjećanjima.“, istaknuo je Ivan Supek, prvi predsjedavajući Kongresa i njegov prvi predsjednik, prigodom otvaranja Kongresa. Bez presedana, taj Kongres održan u Filadelfiji od 21. do 24. svibnja 1976., okupio je značajne svjetske humaniste – nobelovce sa svih kontinenata i, na promišljen način, uspostavio prirodne osnove za pozitivan razvoj suvremene civilizacije, te stvorio nadu svim ljudima svijeta. Kongres je prihvatio Dubrovačku deklaraciju, koja se od tada zove Dubrovačko – filadelfijska deklaracija. Uz tu Deklaraciju, razvijena je i prihvaćena Dubrovačka izjava i Dubrovačka deklaracija o povjerenju. Kongres je utvrdio STATUS Internacionalne lige humanista. Neovisna, neutralna od vlasti, politike, religije. Djeluje u svim državama gdje se ukaže potreba. Internacionalna liga humanista nije ni američka, ni engleska, ni jugoslavenska, niti bilo koje zemlje iz koje su idejni tvorci, već INTERNACIONALNA, sa sjedištem u onoj zemlji koju odredi Kongres ILH i koja ponudi uvjete za rad i djelovanje misije, bez miješanja u misiju. Internacionalna liga humanista ima status međunarodne organizacije od javnog interesa, s ciljem da zastupa interese svih građana i radi na promicanju mira u svijetu. Liga je ustanovljena s namjerom da se planovi za mir ostvare diljem svijeta u njegovom punom smislu. Sve što je potrebno za dostizanje svih ovih svjetskih težnji sažeto je u Dubrovačko – filadelfijskoj deklaraciji i Deklaraciji o povjerenju koje predstavljaju akcelerator evolucije istinske civilizacije.

 
MOTRIŠTA: Kojim ćete aktivnostima ove godine obilježiti 35. obljetnicu Internacionalne lige humanista?

 

ŠURLAN: Planiran je niz aktivnosti na planu obilježavanja 35. obljetnice Internacionalne lige humanista, među kojima ću spomenuti najvažnije:
Održavanje Svjetskog kongresa jedinstva humanista posvećenog dvjema krucijalnim temama današnjice, a to su „Djeca svijeta ranjenog djetinjstva“ (djeca bez djetinjstva) i problemima odgoja i obrazovanja za 21. stoljeće „Obrazovanjem protiv siromaštva – odgojem protiv nasilja do kulture mira“.
Kongres će se održati od 23. do 26. listopada 2009. godine, u Dubrovniku. Za predsjednicu Internacionalnog komiteta majki solidarnosti za djecu svijeta ranjenog djetinjstva, imenovana je gđa Jadranka Kosor. Za predsjednicu Komiteta majki solidarnosti za Republiku Hrvatsku imenovana je prof. Dubravka Šuica. Drugi značajan događaj koji planiramo organizirati krajem rujna 2009. godine jest
Mala olimpijada sporta i kulture mira „Juan Antonio Samaranch“, uz sudjelovanje mladih i sportskih veterana iz država Zapadnog Balkana. Planirano je da mladi budu srednjoškolskog uzrasta. Svaka država predstavit će se s 11 srednjoškolaca, 3 veterana iz nogometa i 3 olimpijske legende. Razumije se da će tu sudjelovati velik broj ostalih stručnjaka i simpatizera mira i sporta.
Treći događaj odnosi se na obilježavanje Dana Internacionalne lige humanista, od 21. do 24. svibnja 2009. godine. Tim povodom, u Dubrovniku će se održati Svečana akademija posvećena miru, zaštiti ljudskih prava i dostojanstva. Na Akademiji će sudjelovati poznate osobe iz oblasti ljudskih prava, slobode i misije mira Internacionalne lige humanista. Sudionicima će se posebno obratiti laureat odvjetnik Josip Muselimović, dobitnik Zlatne povelje „Linus Pauling“ s Plaketom i Značkom, u znak trajne zahvalnosti za njegov humani i etički pristup obrani obespravljenih, pod nazivom „Poštovani suci, zašto?“.
Svečana ceremonija proglašenja g. Josipa Muselimovića Internacionalnim odvjetnikom obespravljenih, uz uručenje ranije dodijeljenih priznanja, održat će se u Mostaru, najvjerojatnije, 7. lipnja. Priznanja će uručiti dr. Elisabeth Rehn, predsjednica Internacionalne lige humanista. Prema propozicijama i Pravilima o nominiranju, valorizaciji nominacije i izboru nije dopušteno davati obrazloženje prije službenog izbora.

 
MOTRIŠTA: Već dulje vrijeme Internacionalna liga humanista istražuje najpovoljnije lokalitete za otvaranje Međunarodnog centra za sport i kulturu mira koji bi nosio ime „Juan Antonio Samaranch“. Imamo nekih naznaka kako bi to mogao biti lokalitet sadašnjeg rekreacijskog centra Buna. O čemu se, zapravo, radi?

 

ŠURLAN: Poznato je da je Peti kongres Internacionalne lige humanista održan u Dubrovniku od 23. do 26. listopada 2005. u znak priznanja za doprinos miru u svijetu za vrijeme njegova mandata na čelu MOK-a, NJ. E. Juana Antonia Samarancha proglasio Humanistom 20. stoljeća i prvim počasnim veleposlanikom sporta i kulture mira, uz najviša humanistička priznanja. Internacionalne liga humanista odobri Ligi korištenje njegovog imena i lika za razvijanje mira, solidarnosti, borbe protiv droge i nasilja u korist stvaranja mostova prijateljstva među mladima diljem svijeta, kroz razne oblike djelovanja s mladima putem razvoja amaterskog sporta.
Na temelju Projekta koji smo uradili na spomenutu temu, NJ.E. Juan Antonio Samaranch dao je EKSKLUZIVNO PRAVO Internacionalnoj ligi humanista da se može koristiti njegovim imenom i likom, na način kako to Liga humanista radi, sa zaštitom i korištenjem lika i imena svojih osnivača. Za zaštitu svoga lika i imena osobno je zadužio glavnoga tajnika Internacionalne lige humanista.
U želji da stvorimo Međunarodni centar za mlade koji bi se odgajali u duhu mira i prijateljstva, pod imenom Juana Antonia Samarancha, od 2006. godine do danas, tražimo lokaciju koja bi zadovoljila kriterije za jedan takav centar. Prema našim propozicijama, najviše bi odgovarao lokalitet Bune, o čemu smo informirali gradonačelnika Mostara, predsjednika Vlade HNK/ž, te našega laureata iz Mostara gosp. Josipa Muselimovića. Hoće li svi ovi i drugi subjekti moći, znati i htjeti prepoznati značenje jednog ovakovog centra za sveukupnu afirmaciju ne samo Mostara već regije, države BiH, pa i zemalja u okruženju, ostaje vidjeti. Oni su na potezu. Poslije njih idemo u Hrvatsku.

 
MOTRIŠTA: Koje nam najznačajnije planove i programe, uz ove već spomenute, možete istaknuti kao prioritetne u ovoj godini?

 

ŠURLAN: Više je aktivnosti koje se odvijaju u okviru Internacionalne lige humanista i pod pokroviteljstvom Lige. Spomenut ću samo neke od njih:
Prvo, ustrojavanje našega Internacionalnog univerziteta Philip Noel-Baker u sveučilišnu ustanovu internacionalnog tipa po Bolonjskoj deklaraciji, s fakultetima za turizam i hotelijerstvo, za informatiku i aktuarstvo, za biznis i menadžment i druge, s vlastitim Institutom za znanstveno – istraživački rad i razvoj. Na ovom planu imamo veliku znanstvenu i pragmatičnu potporu od Sveučilišta u Dubrovniku, koje je prvi potpisnik Sporazuma o suradnji.
Drugi, izuzetno važan projekt jest uključivanje Internacionalne lige humanista u proces održivog povratka u Bosni i Hercegovini, uz primjenu svoga vlastitog projekta „Istinom do održivog povratka“. Naša je želja da se svi vrate na svoja ognjišta, pod uvjetima koji odgovaraju dostojanstvu čovjeka 21. stoljeća.
Treći projekt, koji se nastavlja iz godine u godinu, jest zaštita djece od zloporabe i nasilja, zaštita zdravlja građana, borba protiv droge i nasilja, uključivanje mladih u demokratske procese u BiH, itd.
Četvrta aktivnost koja nas preokupira jest „Humano i etičko upravljanje okolišem“, zaštita prirodno – atraktivnih resursa od zagađenja i uništenja. Otvaranje edukacije za mlade pod nazivom „Humani i etički preduvjeti za izgradnju pozitivnog odnosa prema okolišu“. Odgoj i obrazovanje mladih u osnovnim školama za zdrav okoliš te niz drugih nespomenutih programa i projekata. 

MOTRIŠTA: Poznato je da ste 2008. obilježili 30. obljetnicu kako Liga bira i promovira pojedince, organizacije i institucije iz zemlje i svijeta te dodjeljuje određena priznanja. O kakvim se priznanjima radi, kako se biraju dobitnici i dodjeljuju priznanja?
 

ŠURLAN: Od 1978. do danas, svake godine po posebnim kriterijima, metodologiji i Pravilima, obavlja se nominacija, selekcija i dodjela priznanja laureatima koji su ispunili tražene kriterije iz područja znanosti, kulture, ljudskih prava, socijalne politike, umjetnosti, liderstva i drugih nespomenutih područja. Svaka oblast ima svoje priznanje od Zahvalnice, Diplome, Plakete, Zlatne značke, do najviših priznanja koja nose ime naših utemeljitelja, kao što su Povelja Linus Pauling, Povelja Philip Noel-Baker do počasnih priznanja, poput Ambasador sporta i kulture mira, Internacionalni odvjetnik obespravljenih, državnik desetljeća, stoljeća, i sl.
Od nominacije do izbora ponekad zna proći dosta vremena. Nakon što netko od uglednih osoba (fizička ili pravna) nominira kandidata, tada se, putem elektronske pošte, na adrese 11 – 15 osoba iz više zemalja, koje su iz određenog područja iz kojeg se kandidat nominira, koje ne znaju jedna za drugu, traži mišljenje. Ako od 15 članova odabranih za ocjenu nominacije 14 da pozitivnu ocjenu, kandidat je prošao. Ako se dobije manje od 14 glasova, kandidatura se može ponoviti, ali za iduću godinu i s promijenjenim članovima. Tako od nominacije do izbora, za neke traje izbor dvije godine, tri, pa i četiri godine, dok nisu u prvome krugu dobili glasove. Do sada su Povelju Linus Pauling dobili: Sveti Otac Ivan Pavao II., Juan Antonio Samaranch, dr. Mahathir bin Mogamad (Malezija), acc. Luko Paljetak, acc. Ivan Supek, msgr. dr. franjo Komarica, acc. Seid Huković...

 
MOTRIŠTA: Kako sa stajališta misije Internacionalne lige humanista, njezinih humanih i etičkih načela, ocjenjujete današnje naše društvo?

 

ŠURLAN: Pa, bilo bi bolje da ste pitali g. Josipa Muselimovića, našega v.d. predsjednika Internacionalnog komiteta za humanizam, etiku, zaštitu ljudskih prava i društveni poredak, koji bi Vam najbolje dao ocjenu moralnosti društva. Što kazati? Neizlječiva gangrena „bolesne“ politike i nadobudnih političara zahvatila je naše društvo - od vrha do dna. Posvađana politika uvela nas je u jednu vrstu mirnodopskoga drugog rata najniže razine i grubih udaraca.

Ne puca se, ali se ratuje ubojitim
riječima od kojih se, također, može poginuti. Ušli smo u neku vrstu poslijeratnog rata, koji je veoma okrutan i opasan za izgradnju zdravih društvenih odnosa. Rat i tranzicija su, umnogome, izokrenuli uobičajena pravila društvenih ponašanja i stavili nas na taj  način pred velike izazove i nedoumice. Kako pobijediti Sedam smrtnih grijeha suvremene civilizacije, kako ih je definirao Mahatma Gandi: 1) bogatstvo bez rada; 2) užitak bez savjesti; 3) znanje bez karaktera; 4) posao bez morala; 5) znanost bez čovječnosti; 6) religija bez žrtve; 7) politika bez načela.

 Kako ove smrtne grijehe pobijediti?!

Droga je uslikana i spaljena (Euroblic) 03.08.2018 - autor Nikola Morača

OGNJEN ZEKIĆ, PREDSEDNIK KOMISIJE ZA UNIŠTAVANJE DROGA BIH 

Droga je uslikana i spaljena

 “EuroBlic” je u posedu zapisnika i fotografija sa spaljivanja 700 kilograma narkotika zaplenjenih u policijskim akcijama u proteklih pet godina na području BiH.

 Prema ovim dokumentima kompletnu akciju od utovara, transporta pa do konačnog spaljivanja više stotina kilograma kokaina, heroina, amfetamina, marihuane, odnosno skanka u pećima RiTE „Gacko" pratila je petočlana Komisija Saveta ministara BiH na čelu sa Ognjenom Zekićem koji je bio predsednik. 

Komisija za uništavanje opojnih droga BiH bila je sastavljena od službenika ministarstava pravde, finansija i bezbednosti BiH, koji su fotografisali svaki detalj uništavanja droga.
Podsetimo, prašina oko čitavog slučaja se digla nakon što je dan kasnije u krugu preduzeća RiTE
„Gacko" pronađeno par pakovanja različitih vrsta droge u ukupnoj količini od 346 grama.
- Svih 700 kilograma narkotika je uništeno i sve je urađeno u skladu sa zakonom izjavio je predsednik Komisije za uništavanje opojnih droga Ognjen Zekić dodavši kako stoji iza kompletnog saopštenja koje je Ministarstvo bezbednosti BiH objavilo nakon spaljivanja narkotika.
U saopštenju se između ostalog, navodi da je spaljeno više od 700 kilograma droga, najviše marihuane, te amfetamina, kokaina,heroina i sintetičkih droga.
Akcija prikuplja nja narkotika trajala je tri meseca, a ulična vrednost uništenih narkotika procenjena je na oko sedam miliona maraka navedeno je u saopštenju Ministarstva
bezbednosti BiH nakon uništavanja narkotika.
Prema dokumentaciji koja je u posedu “EuroBlica’ 700 kilograma narkotika, koji su bili predmet
uništavanja, jeste iz pravosnažnih presuda Suda BiH, Vrhovnog suda Federacije BiH, kantonalnih sudova u Sarajevu i Novom Travniku, te opštinskih sudova u Kaknju, Tešnju, Bugojnu,Velikoj Kladuši, Zenici, Kiseljaku, Cazinu i Bosanskoj Krupi.
Osim petočlane Komisije, akciju uništavanja pratila su četvorica službenika SIPA,
kao i pripadnici jedinice za specijalnu podršku SIPA, te trojica radnika RiTE „Gacko".
Droga je nakon pregleda i vaganja bila zapakovana u 161 plastičnu vreću i kutiju koje su zapeča
ćene plastičnim plombama i numerisane.
Transport droge izvršili su pripadnici SIPA, a članovi Komisije za uništavanje opojnih droga su
izvršili uvid u nepovredivost vreća i kutija u koje je bila zapakovana droga navodi se u izveštaju Komisije, čime se u potpunosti negira bilo kakva veza sa drogom koja je u krugu preduzeća RiTE „Gacko"pronađena dan kasnije.
Prilikom preuzimanja droge, veštaci Centra za forenzičku i informatičku podršku Federalne
uprave policije su, prema zahtevu članova Komisije, vršili preliminarna testiranja, te
izuzeli uzorke radi veštačenja navodi se u izveštaju Komisije koji je u posedu “EuroBlica”.
Veštačenje sporne droge koja je pronađena u krugu preduzeća trebalo bi otkloniti
sve sumnje.

 

 

“Establishing the South East Europe Women Police Officers Network” - agenda for the presentation of the report (28th October 2010)

 

Complementary 1325 Global Open Day Events in Bosnia and Herzegovina and Serbia - report (RWL - June 7, 2010)

Summary 

(the full report below as pdf) 

In support of the UN-led Global Open Day for 1325 events, the Regional Women's Lobby decided to organize a set of complementary meetings in two of the countries where the Lobby is active: Serbia and Bosnia and Herzegovina. As neither of these countries has a DPKO mission the UN team leading the Global Open Day initiative was not planning on promoting the event in these locations. RWL decided to use this space to organize its own meetings, complementary to the Global Open Day, with the UN RC and other stakeholders to brief them and discuss 1325 implementation issues. Both of these countries has a specific set of contextual factors influencing UNSCR 1325 implementation and this background will be discussed below. It should be mentioned that the RWL also planned to meet with the President of Croatia, however due to scheduling factors beyond their control the meeting was cancelled. As an alternative, the RWL members will meet with the President of Croatia and the UN RC in Croatia at the end of June after the RWL Regional 1325 +10 conference to brief them on the state of 1325 implementation in Croatia and the region.  

Context: Bosnia and Herzegovina 

Bosnia and Herzegovina is a post-conflict country that endured years of bloody conflict in the 1990s in which rape was used a method of warfare and women were victims of shockingly brutal gender based violence. The conflict led to the militarization of these societies and women were pushed out of public spaces and lost many of the social and economic rights they had under the socialist regime. In terms of Protection Laws and Mechanisms, Bosnia has a good legal framework. There is a Law on Gender Equality; a legislated electoral quota; a Law on Domestic Violence; a NAP on Gender Equality; a NAP for Domestic Violence; and a NAP on implementation 1325 is in the process of being adopted by Parliament. Mechanisms for implementation of these laws include a government Agency for Gender Equality, Gender Centers at the "entity" level of Bosnia's structure and a municipal/regional gender focal point network. In terms of political participation of women there is 14% women in parliament and only 2-3 women in executive decision-making positions at any point in time. In the security sector there are 2% women in the military at higher ranks, 5% at lower ranks and 5% at enlisted level. In the police there are 8% women. Civil society is very active in Bosnia on monitoring and advocating for 1325, with the strong support of UNIFEM. For example, the NGO Zene zenama has been active in raising awareness of 1325. They have translated the resolution into the local language and dispersed it and united with national and international security sector organizations in a program entitled "Participation of the public in security: UNSCR 1325 in BiH". Examples of goals of the 1325 program are the creation of a women's police network at the national level and the establishment of cooperation between women's NGOs and EUFOR local observation teams on women's human rights and security. Civil society in Bosnia has also been involved in the NAP drafting process and will take on promotion of the NAP in local communities along with representatives of ministries, entities and cantons.  

Context: Serbia 

Serbia is a post-conflict society with a high degree of militarization. Serbia experienced the effects of the wars in the region first indirectly, and then directly with the NATO bombing in 1999. As in Bosnia, women in Serbia were also pushed out of public spaces and decision-making during the conflicts of the 1990s as public attitudes reinforced the view that women were not the decisive politicians or strong leaders needed during the wars and conflicts that Serbia was involved in Croatia, Bosnia and Kosovo. Serbia, like Bosnia, also has an extensive legal framework related to gender equality. There is a Law on Gender Equality; a legislated electoral quota; an Anti-discrimination Law; a NAP on Gender Equality; and a NAP on implementation 1325 is currently being drafted by a working group of stakeholders. The gender mechanisms include a Department for Gender Equality within the Ministry of Labour and Social Affairs and a network of gender focal points at the regional level of government. Women's participation in parliament is about 21% and at the executive level 5 of 27 ministers are currently women. In the military 0.5% of officer are women and 5% of enlisted soldiers are women. The NGO Women in Black has been advocating for a women's perspective of security since the very beginning of the conflicts in the region in the 1990s and has been active in promoting peace and security issues from a feminist perspective and from an anti-nationalist and anti-militaristic perspective. Needless to say their consistent opposition to the regime was not without danger to themselves, especially in the charged atmosphere of the 1990s.

Mogherini i Pendeš potpisale Okvirni sporazum o učešću BiH u operacijama EU za upravljanje kriznim situacijama (15.09.2015.)

 https://www.facebook.com/permalink.php?story_fbid=946923118684200&id=163322863710900

(objavljeno 15.09.2015. na Facebook stranici Evropske unije u Bosni i Hercegovini) 

 

Visoka predstavnica i potpredsjednica Europske komisije Federica Mogherini i ministrica odbrane Bosne i Hercegovine (BiH), Marina Pendeš, potpisale su danas u Briselu Okvirni sporazum o učešću Bosne i Hercegovine u operacijama Europske unije za upravljanje kriznim situacijama.

Sporazumom se utvrđuje pravni okvir za moguće buduće učešće Bosne i Hercegovine u cijelom nizu vojnih operacija i civilnih misija za upravljanje krizama, koje vodi Europska unija. Nadalje, Sporazum predstavlja korak naprijed ka većoj razini strukturirane saradnje između Europske unije i Bosne i Hercegovine u oblasti sigurnosti.

Europska unija već ima bliske veze sa Bosnom i Hercegovinom u ovoj oblasti, imajući u vidu da je EUPM, kao prva policijska misija u okviru zajedničke sigurnosne i odbrambene politike Europske unije, pokrenuta u Bosni i Hercegovini 2003. godine, sa zadatkom pružanja podrške policijskim strukturama do juna 2012. godine. U zemlji je još uvijek prisutna vojna operacija Europske unije pod nazivom EUFOR Althea, koja je pokrenuta 2004. godine i koja podržava napore BiH na očuvanju sigurnog okruženja kao i propratnu izgradnju kapaciteta i obuku Ministarstva obrane BiH i Oružanih snaga BiH.

Novi okvirni instrument će omogućiti, ovisno od slučaja do slučaja, nesmetano uključivanje Bosne i Hercegovine u postojeće i buduće napore Europske unije na upravljanju krizama širom svijeta. Time će se izbjeći nepotrebna odlaganja, u slučajevima gdje je Bosna i Hercegovina pozvana i saglasna da učestvuje u operacijama Europske unije. Instrument će također ponuditi i bolju učinkovitost i fleksibilnost u odgovoru na buduće krize te dodatno potvrđuje sve veću ulogu i zalaganje BiH da djeluje kao faktor sigurnosti, posebno u svjetlu njenog trenutačnog doprinosa mirovnim operacijama UN-a.

 

 

America used Islamists to arm the Bosnian Muslims (The Guardian, 22 April, 2002) by Richard J Aldrich

 https://amp.theguardian.com/world/2002/apr/22/warcrimes.comment?fbclid=IwAR2n8L3llMx7R7B8q0rKYq4YLJUEaWfYIhUpwviBqowNMXa9x8UEv3GxQHk

 

 The Srebrenica report reveals the Pentagon's role in a dirty war

The official Dutch inquiry into the 1995 Srebrenica massacre, released last week, contains one of the most sensational reports on western intelligence ever published. Officials have been staggered by its findings and the Dutch government has resigned. One of its many volumes is devoted to clandestine activities during the Bosnian war of the early 1990s. For five years, Professor Cees Wiebes of Amsterdam University has had unrestricted access to Dutch intelligence files and has stalked the corridors of secret service headquarters in western capitals, as well as in Bosnia, asking questions.

His findings are set out in "Intelligence and the war in Bosnia, 1992-1995". It includes remarkable material on covert op
erations, signals interception, human agents and double-crossing by dozens of agencies in one of dirtiest wars of the new world disorder. Now we have the full story of the secret alliance between the Pentagon and radical Islamist groups from the Middle East designed to assist the Bosnian Muslims - some of the same groups that the Pentagon is now fighting in "the war against terrorism". Pentagon operations in Bosnia have delivered their own "blowback".

In the 1980s Washington's secret services had assisted Saddam Hussein in his war against Iran. Then, in 1990, the US fought him in the Gulf. In both Afghanistan and the Gulf, the Pentagon had incurred debts to Islamist groups and their Middle Eastern sponsors. By 1993 these groups, many supported by Iran and Saudi Arabia, were anxious to help Bosnian Muslims fighting in the former Yugoslavia and called in their debts with the Americans. Bill Clinton and the Pentagon were keen to be seen as creditworthy and repaid in the form of an Iran-Contra style operation - in flagrant violation of the UN security council arms embargo against all combatants in the former Yugoslavia.

 

The result was a vast secret conduit of weapons smuggling though Croatia. This was arranged by the clandestine agencies of the US, Turkey and Iran, together with a range of radical Islamist groups, including Afghan mojahedin and the pro-Iranian Hizbullah. Wiebes reveals that the British intelligence services obtained documents early on in the Bosnian war proving that Iran was making direct deliveries.

Arms purchased by Iran and Turkey with the financial backing of Saudi Arabia made their way by night from the Middle East. Initially aircraft from Iran Air were used, but as the volume increased they were joined by a mysterious fleet of black C-130 Hercules aircraft. The report stresses that the US was "very closely involved" in the airlift. Mojahedin fighters were also flown in, but they were reserved as shock troops for especially hazardous operations.

Light weapons are the familiar currency of secret services seeking to influence such conflicts. The volume of weapons flown into Croatia was enormous, partly because of a steep Croatian "transit tax". Croatian forces creamed off between 20% and 50% of the arms. The report stresses that this entire trade was clearly illicit. The Croats themselves also obtained massive quantities of illegal weapons from Germany, Belgium and Argentina - again in contravention of the UN arms embargo. The German secret services were fully aware of the trade.

Rather than the CIA, the Pentagon's own secret service was the hidden force behind these operations. The UN protection force, UNPROFOR, was dependent on its troop-contributing nations for intelligence, and above all on the sophisticated monitoring capabilities of the US to police the arms embargo. This gave the Pentagon the ability to manipulate the embargo at will: ensuring that American Awacs aircraft covered crucial areas and were able to turn a blind eye to the frequent nightime comings and goings at Tuzla.

Weapons flown in during the spring of 1995 were to turn up only a fortnight later in the besieged and demilitarised enclave at Srebrenica. When these shipments were noticed, Americans pressured UNPROFOR to rewrite reports, and when Norwegian officials protested about the flights, they were reportedly threatened into silence.

Both the CIA and British SIS had a more sophisticated perspective on the conflict than the Pentagon, insisting that no side had clean hands and arguing for caution. James Woolsey, director of the CIA until May 1995, had increasingly found himself out of step with the Clinton White House over his reluctance to develop close relations with the Islamists. The sentiments were reciprocated. In the spring of 1995, when the CIA sent its first head of station to Sarajevo to liaise with Bosnia's security authorities, the Bosnians tipped off Iranian intelligence. The CIA learned that the Iranians had targeted him for liquidation and quickly withdrew him.

Iranian and Afghan veterans' training camps had also been identified in Bosnia. Later, in the Dayton Accords of November 1995, the stipulation appeared that all foreign forces be withdrawn. This was a deliberate attempt to cleanse Bosnia of Iranian-run training camps. The CIA's main opponents in Bosnia were now the mojahedin fighters and their Iranian trainers - whom the Pentagon had been helping to supply months earlier.

 

Meanwhile, the secret services of Ukraine, Greece and Israel were busy arming the Bosnian Serbs. Mossad was especially active and concluded a deal with the Bosnian Serbs at Pale involving a substantial supply of artillery shells and mortar bombs. In return they secured safe passage for the Jewish population out of the besieged town of Sarajevo. Subsequently, the remaining population was perplexed to find that unexploded mortar bombs landing in Sarajevo sometimes had Hebrew markings.

The broader lessons of the intelligence report on Srebrenica are clear. Those who were able to deploy intelligence power, including the Americans and their enemies, the Bosnian Serbs, were both able to get their way. Conversely, the UN and the Dutch government were "deprived of the means and capacity for obtaining intelligence" for the Srebrenica deployment, helping to explain why they blundered in, and contributed to the terrible events there.

 

Secret intelligence techniques can be war-winning and life-saving. But they are not being properly applied. How the UN can have good intelligence in the context of multinational peace operations is a vexing question. Removing light weapons from a conflict can be crucial to drawing it down. But the secret services of some states - including Israel and Iran - continue to be a major source of covert supply, pouring petrol on the flames of already bitter conflicts.

· Richard J Aldrich is Professor of Politics at the University of Nottingham. His 'The Hidden Hand: Britain, America and Cold War Secret Intelligence' is published in paperback by John Murray in August.

richard.aldrich@nottingham.ac.uk

 

Recommending items to more than a billion people (by Aleksandar Ilic and Maja Kabiljo POSTED ON JUNE 2, 2015 TO Core Data, ML Applications)

 https://engineering.fb.com/2015/06/02/core-data/recommending-items-to-more-than-a-billion-people/

 

POSTED ON JUNE 2, 2015 TO Core Data, ML Applications

Recommending items to more than a billion people

By Aleksandar Ilic, Maja Kabiljo

The growth of data on the web has made it harder to employ many machine learning algorithms on the full data sets. For personalization problems in particular, where data sampling is often not an option, innovating on distributed algorithm design is necessary to allow us to scale to these constantly growing data sets.

Collaborative filtering (CF) is one of the important areas where this applies. CF is a recommender systems technique that helps people discover items that are most relevant to them. At Facebook, this might include pages, groups, events, games, and more. CF is based on the idea that the best recommendations come from people who have similar tastes. In other words, it uses historical item ratings of like-minded people to predict how someone would rate an item.

CF and Facebook scale

Facebook’s average data set for CF has 100 billion ratings, more than a billion users, and millions of items. In comparison, the well-known Netflix Prize recommender competition featured a large-scale industrial data set with 100 million ratings, 480,000 users, and 17,770 movies (items). There has been more development in the field since then, but still, the largest numbers we’ve read about are at least two orders of magnitude smaller than what we’re dealing with.

A challenge we faced is to design a distributed algorithm that is going to scale to these massive data sets, and how to overcome issues that arose because of certain properties of our data (like skewed item degree distribution, or implicit engagement signals instead of ratings).

As we’ll discuss below, approaches used in existing solutions would not efficiently handle our data sizes. Simply put, we needed a new solution. We’ve written before about Apache Giraph, a powerful platform for distributed iterative and graph processing, and the work we put into making it scale to our needs. We’ve also written about one of the applications we developed on top of it about graph partitioning. Giraph works extremely well on massive data sets, it is easily extensible, and we have a lot of experience in developing highly performant applications on top of it. Therefore, Giraph was our obvious choice for this problem.

Matrix factorization

A common approach to CF is through matrix factorization, in which we look at the problem as having a set of users and a set of items, and a very sparse matrix that represents known user-to-item ratings. We want to predict missing values in this matrix. In order to do this, we represent each user and each item as a vector of latent features, such that dot products of these vectors closely match known user-to-item ratings. The expectation is that unknown user-to-item ratings can be approximated by dot products of corresponding feature vectors, as well. The simplest form of objective function, which we want to minimize, is:

Here, r are known user-to-item ratings, and x and y are the user and item feature vectors that we are trying to find. As there are many free parameters, we need the regularization part to prevent overfitting and numerical problems, with gamma being the regularization factor.

It is not currently feasible to find the optimal solution of the above formula in a reasonable time, but there are iterative approaches that start from random feature vectors and gradually improve the solution. After some number of iterations, changes in feature vectors become very small, and convergence is reached. There are two commonly used iterative approaches.

Stochastic gradient descent optimization

Stochastic gradient descent (SGD) optimization was successfully practiced in many other problems. The algorithm loops through all ratings in the training data in a random order, and for each known rating r, it makes a prediction r* (based on the dot product of vectors x and y) and computes prediction error e. Then we modify x and y by moving them in the opposite direction of the gradient, yielding certain update formulas for each of the features of x and y.

Alternating least square

Alternating least square (ALS) is another method used with nonlinear regression models, when there are two dependent variables (in our case, vectors x and y). The algorithm fixes one of the parameters (user vectors x), while optimally solving for the other (item vectors y) by minimizing the quadratic form. The algorithm alternates between fixing user vectors and updating item vectors, and fixing item vectors and updating user vectors, until the convergence criteria are satisfied.

Standard approach and problems

In order to efficiently solve the above formula in a distributed way, we first looked at how systems that are similar in design to Giraph do it (using message passing instead of map/reduce). The standard approach corresponds to having both users and items as vertices of a graph, with edges representing known ratings. An iteration of SGD/ALS would then send user and/or item feature vectors across all the edges of the graph and do local updates.

There are a few problems with this solution:

1. Huge amount of network traffic: This is the main bottleneck of all distributed matrix factorization algorithms. Since we send a feature vector across each edge of the graph, the amount of data sent over the wire in one iteration is proportional to #Ratings * #Features (here and later in the text we use # as notation for ‘number of’). For 100 billion ratings and 100 double features, this results in 80 TB of network traffic per iteration. Here we assumed that users and items are distributed randomly, and we are ignoring the fact that some of the ratings can live on the same worker (on average, this should be multiplied by the factor 1 – (1 / #Workers)). Note that smart partitioning can’t reduce network traffic by a lot because of the items that have large degrees, and that would not solve our problem.
2. Some items in our data sets are very popular, so item degree distribution is highly skewed: This can cause memory problems — every item is receiving degree * #Features amount of data. For example, if an item has 100 million known ratings and 100 double features are used, this item alone would receive 80 GB of data. Large-degree items also cause processing bottlenecks (as every vertex is atomically processed), and everyone will wait for a few largest-degree items to be finished.
3. This does not implement SGD exactly in the original formulation: Every vertex is working with feature vectors that it received in the beginning of the iteration, instead of the latest version of them. For example, say item A has ratings for users B and C. In a sequential solution, we’d update A and B first, getting A’ and B’, and then update A’ and C. With this solution, both B and C will be updated with A, the feature vector for the item from the beginning of the iteration. (This is practiced with some lock-free parallel execution algorithms and can slow down the convergence.)

Our solution — rotational hybrid approach

The main problem is sending all updates within each iteration, so we needed a new technique of combining these updates and sending less data. First we tried to leverage aggregators and use them to distribute item data, but none of the formulas we tried for combining partial updates on the item feature vectors worked well.

We finally came up with an approach that required us to extend Giraph framework with worker-to-worker messaging. Users are still presented as the vertices of the graph, but items are partitioned in #Workers disjoint parts, with each of these parts stored in global data of one of the workers. We put all workers in a circle, and rotate the items in clockwise direction after each superstep, by sending worker-to-worker messages containing items from each worker to the next worker in the line.

This way, in each superstep, we process part of the worker’s user ratings for the items that are currently on the worker, and therefore process all ratings after #Workers supersteps. Let’s analyze the issues the previous solutions had:

1. Amount of network traffic: For SGD, the amount of data sent over the wire in one iteration is proportional to #Items * #Features * #Workers, and it doesn’t depend on the number of known ratings anymore. For 10 million items, 100 double features, and 50 workers, this brings a total of 400 GB, which is 20x smaller than in the standard approach. Therefore, for #Workers <= #Ratings / #Items rotational approach performs much better, i.e., if the number of workers is less than the average item degree. In all data sets that we are using, the items with small degree are ignored from consideration, as those do not represent good recommendations and can be just noise, so average item degree is large. We’ll talk below more about ALS.
2. Skewed item degrees: This is no longer a problem — user vertices are the only ones doing processing, and items never hold information about their user ratings.
3. Computation of SGD: This is equal as in a sequential solution, because there is only one version of a feature vector at any point of time, instead of having copies of them sent to many workers and doing updates based on that.

The computation with ALS is trickier than with SGD, because in order to update a user/item, we need all its item/user feature vectors. The way updates in ALS actually go is that we are solving a matrix equation of type A * X = B, where A is #Features x #Features matrix and B is 1 x #Features vector, and A and B are calculated based on user/item feature vectors forming all known ratings for item/user. So when updating items, instead of rotating just their feature vectors, we can rotate A and B, update them during each of #Workers supersteps and calculate new feature vectors in the end. This increases the amount of network traffic to #Items * #Features² * #Workers. Depending on proportions between all the data dimensions, for some items this is better than the standard approach, and for some it isn’t.

This is why a blend of our rotational approach and the standard approach gives the superior solution. By looking at item with some degree, in the standard approach the amount of network traffic associated with it is degree * #Features, and with our rotational approach, it’s #Workers * #Features². We’ll still update items in which degree < #Workers * #Features using the standard approach, and we’ll use our rotational approach for all larger-degree items, and therefore significantly improve performance. For example, for 100 double features and 50 workers, the item degree limit for choosing an approach is around 5,000.

To solve the matrix equation A * X = B we need to find the inverse A^-1, for which we use open source library JBLAS, which had the most efficient implementation for the matrix inverse.

As SGD and ALS share the same optimization formula, it is also possible to combine these algorithms. ALS is computationally more complex than SGD, and we included an option to do a combination of some number of iterations of SGD, followed by a single iteration of ALS. For some data sets, this was shown to help in the offline metrics (e.g., root mean squared error or mean average rank).

We were experiencing numerical issues with large-degree items. There are several ways of bypassing this problem (ignoring these items or sample them), but we were using regularization based on the item and user degrees. That keeps the values for user and item vectors in a certain numerical range.

Evaluation data and parameters

In order to measure the quality of recommendations, before running an actual A/B test, we can use a sample of the existing data to compute some offline metrics about how different our estimations are from the actual user preferences. Both of the above algorithms have a lot of hyperparameters to tune via cross-validation in order to get the best recommendations, and we provide other options like adding user and item biases.

The input ratings can be split in two data sets (train and test) explicitly. This can be very useful in cases in which testing data is composed of all user actions in the time interval after all training instances. Otherwise, to construct the test data, we randomly selected T=1 items per user, and keep them apart from training.

During the algorithm, for a certain percent of users we rank all unrated items (i.e., items that are not in the training set) and observe where training and testing items are in the ranked list of recommendations. Then we can evaluate the following metrics: mean average rank (the position in the ranked list, averaged over all test items), precision at positions 1/10/100, mean of the average precision across all test items (MAP), etc. Additionally we compute root mean squared error (RMSE), which amplifies the contributions of the absolute errors between the predictions and the true values. To help monitor convergence and quality of results, after each iteration we are printing all these metrics.

On a sample data set with 35 billion weighted training ratings and 0.2 billion testing ratings, the following figure shows how RMSE reduces on training and testing sets for #Features=8 or #Features=128, where other parameters are fixed.

Item recommendation computation

In order to get the actual recommendations for all users, we need to find items with highest predicted ratings for each user. When dealing with the huge data sets, checking the dot product for each (user, item) pair becomes unfeasible, even if we distribute the problem to more workers. We needed a faster way to find the top K recommendations for each user, or a good approximation of it.

One possible solution is to use a ball tree data structure to hold our item vectors. A ball tree is a binary tree where leafs contain some subset of item vectors, and each inner node defines a ball that surrounds all vectors within its subtree. Using formulas for the upper bound on the dot product for the query vector and any vector within the ball, we can do greedy tree traversal, going first to the more promising branch, and prune subtrees that can’t contain the solution better than what we have already found. This approach showed to be 10-100x faster than looking into each pair, making search for recommendations on our data sets finish in reasonable time. We also added an option to allow for specified error when looking for top recommendations to speed up calculations even more.

Another way the problem can be approximately solved is by clustering items based on the item feature vectors — which reduces the problem to finding top cluster recommendations and then extracting the actual items based on these top clusters. This approach speeds up the computation, while slightly degrading the quality of recommendations based on the experimental results. On the other hand, the items in a cluster are similar, and we can get a diverse set of recommendations by taking a limited number of the items from each cluster. Note that we also have k-means clustering implementation on top of Giraph, and incorporating this step in the calculation was very easy.

Comparison with MLlib

Spark MLlib is a very popular machine-learning library that contains one of the leading open source implementations in this domain. In July 2014, the Databricks team published performance numbers of their ALS implementation on Spark. Experiments were conducted on scaled copies of the Amazon reviews data set, which originally contained 35 million ratings and ran for five iterations.

In the following graph, we compared our rotational hybrid approach (which we implemented in Giraph) with the standard approach (implemented in Spark MLlib, including some additional optimizations, like sending a feature vector at most once to a machine), on the same data set. Due to hardware differences (we had about twice the processing power per machine), in order to make a fair comparison we were looking at total CPU minutes. Rotational hybrid solution was about 10x faster.

Additionally, the largest data set on which experiments were conducted with standard approach had 3.5 billion ratings. With rotational hybrid approach, we can easily handle more than 100 billion ratings. Note that quality of results is the same for both, and all performance and scalability gains come from different data layout and decreased network traffic.

Facebook use cases and implicit feedback

We used this algorithm for multiple applications at Facebook, e.g. for recommending pages you might like or groups you should join. As already mentioned, our data sets are composed of more than 1 billion users and usually tens of millions of items. There are actually many more pages or groups, but we limit ourselves to items that pass a certain quality threshold — where the simplest version is to have the item degree greater than 100. (Fun side note: On the other side, we have some very large-degree pages — the “Facebook for Every Phone” page is actually liked by almost half of Facebook’s current monthly active users.)

Our first iterations included page likes/group joins as positive signals. The negative signals on Facebook are not as common (negative signals include unliking a page or leaving a group after some time). Also this may not actually mean that a user has negative feedback for that item; instead, he or she might have lost interest in the topic or in receiving updates. In order to get good recommendations, there is a significant need for adding negative items from the unrated pairs in the collection. Previous approaches include randomly picking negative training sample from unrated items (leading to a biased, non-optimal solution) or treating all unknown ratings as negative, which tremendously increases complexity of the algorithm. Here, we implemented adding random negative ratings by taking into account the user and item degrees (adding negative ratings proportional to the user degree based on the item degree distribution), and weighing negative ratings less than positive ones, as we failed to learn a good model with uniform random sampling approach.

On the other hand, we have implicit feedback from users (whether the user is actively viewing the page, liking, or commenting on the posts in the group). We also implemented a well-known ALS-based algorithm for implicit feedback data sets. Instead of trying to model the matrix of ratings directly, this approach treats the data as a combination of binary preferences and confidence values. The ratings are then related to the level of confidence in observed user preferences, rather than explicit ratings given to items.

After running the matrix factorization algorithm, we have another Giraph job of actually computing top recommendations for all users.

The following code just shows how easy it is to use our framework, tune parameters, and plug in different data sets:

CFTrain(
    ratings=CFRatings(table='cf_ratings'),
    feature_vectors=CFVectors(table='cf_feature_vectors'),
    features_size=128,
    iterations=100,
    regularization_factor=0.02,
    num_workers=5,
)

CFRecommend(
    ratings=CFRatings(table='cf_ratings'),
    feature_vectors=CFVectors(table='cf_feature_vectors'),
    recommendations=CFRecommendations(table='cf_recommendations'),
    num_recommendations=50,
    num_workers=10,
)

Furthermore, one can simply implement other objective functions (such as rank optimizations or neighboring models) by extending SGD or ALS computation.

Scalable CF

Recommendation systems are emerging as important tools for predicting user preferences. Our framework for matrix factorization and computing top user recommendations is able to efficiently handle Facebook’s massive data sets with 100 billion ratings. It is easy to use and extend with other approaches.

We are thinking about many improvements and algorithms, including:

• Incorporating the social graph and user connections for providing a better set of recommendations
• Starting from the previous models instead of random initialization, for recurrent learning
• Automatic parameter fitting with cross-validation for optimizing the different metrics for a given data set
• Trying out better partitioning and skipping machines that don’t need certain item data during rotations

We are actively working on recommendations and many other applications on top of Giraph, so stay tuned for more exciting features and development in this field.

Thanks to Dionysios Logothetis, Avery Ching, Sambavi Muthukrishnan and Sergey Edunov from the Giraph team who made this work possible and helped write this story, and Liang Xiong and Bradley Green for early experimentation and all feedback and insights.