路英川,李 鵬,王 浩,張 翔,湯宇磊,謝 亙
大數(shù)據(jù)時代礦床學(xué)研究發(fā)展?fàn)顩r綜述
路英川1,李 鵬1,王 浩1,張 翔1,湯宇磊1,謝 亙2
(1. 中國地質(zhì)調(diào)查局地球物理調(diào)查中心,河北 廊坊 065000;2. 中國地質(zhì)調(diào)查局廊坊自然資源綜合調(diào)查中心,河北 廊坊 065000)
近年來礦床學(xué)的研究出現(xiàn)了如成礦模式創(chuàng)新性不足以及成礦系列與成礦規(guī)律視角單一等瓶頸。回顧礦床學(xué)的發(fā)展歷程表明,礦床學(xué)研究的每一次突破和飛躍,都與新科學(xué)技術(shù)的發(fā)展息息相關(guān)。隨著“大數(shù)據(jù)”“智能化”時代的來臨,人工智能大數(shù)據(jù)深度學(xué)習(xí)等新技術(shù)正在蓬勃發(fā)展。地質(zhì)數(shù)據(jù)具有“大數(shù)據(jù)”的大量性、高速性、多樣性和價值性的“4V”的特征,還有多元性、多維性、多源性、異構(gòu)性、時空性等特點。通過對近十余年國內(nèi)外相關(guān)文獻(xiàn)的統(tǒng)計對比分析,對人工智能、機器學(xué)習(xí)、深度學(xué)習(xí)等之間的從屬關(guān)系和主要特征進(jìn)行了闡述,對隨機森林算法、卷積神經(jīng)網(wǎng)絡(luò)、決策樹算法、樸素貝葉斯和支持向量機等算法在礦床研究中的實例進(jìn)行了梳理,認(rèn)為人工智能技術(shù)引領(lǐng)全球礦產(chǎn)資源智能勘查研究將成為礦床學(xué)研究發(fā)展的必然方向。
礦床學(xué);成礦模式;大數(shù)據(jù);機器學(xué)習(xí);人工智能
隨著生產(chǎn)力的不斷發(fā)展,礦產(chǎn)資源已經(jīng)逐漸成為人類社會賴以生存和發(fā)展的基礎(chǔ),礦床學(xué)研究也經(jīng)歷一個歷史發(fā)展過程。人們早期依賴肉眼觀察,主要對單一礦種和單一礦床進(jìn)行描述性觀測研究、伴隨地球化學(xué)、地球物理、遙感影像等技術(shù)的普遍應(yīng)用,特別是板塊學(xué)說的引入以及相關(guān)“成礦模式”、“成礦體系”等概念的提出,極大地提高了多種礦床成礦作用與成礦機制的科學(xué)研究水平,也使超大型礦床理論研究取得了不菲的成就[1]。
20世紀(jì)40年代以來,Warren McCulloch和Walter Pitts (1943) 便對人工智能進(jìn)行了嘗試性探索,第一次提出人工神經(jīng)元的數(shù)學(xué)模型[2];此后,Donald Hebb (1949)、Rosenblatt (1958)、Minsky和Papert(1969)以及Rumelhar和Hinton (1986) 等學(xué)者為人工智能技術(shù)的發(fā)展奠定了基礎(chǔ)[3-6]。Mayer-Sch?nhberger Viktor和Cukier Kenneth率先提出了互聯(lián)網(wǎng)信息“大數(shù)據(jù)”的概念[7],并在2013年將《大數(shù)據(jù)時代》這一重要著作呈現(xiàn)到人們面前[8],初步闡述了利用“大數(shù)據(jù)”分析處理取代隨機分析法,從而也可為礦床學(xué)的研究提供新的機遇。
地球科學(xué)與人工神經(jīng)網(wǎng)絡(luò)從雙軌并行到交叉融合的時代已悄然而至。筆者簡要總結(jié)了礦床學(xué)發(fā)展概況,提出礦床研究中存在的成礦模式難以創(chuàng)新、成礦系列與成礦規(guī)律視角單一等問題,并總結(jié)地質(zhì)數(shù)據(jù)具備大量性、高速性、多樣性、價值性的“大數(shù)據(jù)”特點和機器學(xué)習(xí)的常用算法。希望能引發(fā)讀者對大數(shù)據(jù)與地質(zhì)學(xué),特別是人工智能技術(shù)下礦床學(xué)研究發(fā)展方向的思考。
礦床學(xué)是研究礦床成因,揭露成礦規(guī)律,引導(dǎo)礦產(chǎn)勘查的一門地質(zhì)學(xué)分支學(xué)科。礦床學(xué)以“探討成礦機制,指導(dǎo)找礦勘查”為學(xué)科使命,與礦物學(xué)、巖石學(xué)、地球化學(xué)、構(gòu)造地質(zhì)學(xué)等學(xué)科一脈相承[9]。地質(zhì)研究者每次研究思路的改變,將很大程度促進(jìn)礦床學(xué)的發(fā)展。概括來講,礦床學(xué)的研究思路經(jīng)歷了從單一礦種、單一礦床的局部研究到對成礦系列、成礦系統(tǒng)的整體認(rèn)識過程[10]。
20世紀(jì)初期,礦床學(xué)的研究還只是局限于礦床地質(zhì)的探討、成礦熱液的來源、礦田構(gòu)造的展布以及礦床成因的揭露等方面;至20世紀(jì)中后期,伴隨著同位素技術(shù)的引進(jìn)以及板塊理論的普及,極大提高了礦床學(xué)研究的科學(xué)理論水平,并衍生出諸如板塊構(gòu)造與成礦、同位素年代學(xué)以及礦床地球化學(xué)等多個分支學(xué)科;20世紀(jì)80年代以后,關(guān)于超大型礦床、海底熱液成礦、以及地幔柱與成礦的報道紛至沓來[10-13],礦床學(xué)的發(fā)展也進(jìn)入了一個全新的階段。
自從成礦系列的概念被首次提出以來,距今已有百年的研究歷史[10]。成礦系列突破了對礦床的單體研究的局限,更加注重區(qū)域地質(zhì)背景和環(huán)境與成礦作用之間的關(guān)系,將地質(zhì)研究者引向了多礦種共生和多類型礦床共生的廣闊視野中來[10]。“成礦系統(tǒng)”理論則統(tǒng)籌了礦床發(fā)生-形成-保存-破壞等演化歷史與空間分布的統(tǒng)一性,不但進(jìn)一步拓寬了礦床學(xué)研究的方向[12],也完美詮釋了成礦系統(tǒng)與成礦系列的關(guān)系[14]。地質(zhì)學(xué)者們通過對成礦系統(tǒng)的深入研究,取得了豐碩的研究成果[12,15-20]。
探索成礦機制并建立成礦模式一直以來就是礦床學(xué)研究的首要任務(wù),也是指導(dǎo)找礦勘查的理論基礎(chǔ)[9, 21]。通過礦床和成礦帶兩種尺度均可建立相應(yīng)的成礦模式:前者一般是通過對構(gòu)造巖漿控礦、成礦流體演化、礦床地質(zhì)特征等內(nèi)容的研究來限定成礦動力、物質(zhì)來源和成礦過程[9];后者則多集中于不同礦床地質(zhì)和地球化學(xué)特征之間的對比,從而得出區(qū)域成礦模式并指導(dǎo)礦產(chǎn)勘查[9,21]。
同時,礦床勘查方法的革新,觀測手段和測試方法的改進(jìn),推動了礦床成礦元素的賦存狀態(tài)和富集機制研究成果的不斷涌現(xiàn)。
礦床學(xué)的發(fā)展歷程并不是一路平坦,在各個歷史時期都遇到過或大或小的瓶頸。每次瓶頸期的出現(xiàn),都與當(dāng)時其他學(xué)科的科技水平息息相關(guān)。礦床學(xué)的每一次突破和飛躍,也與新科學(xué)技術(shù)的發(fā)展相互聯(lián)系。進(jìn)入21世紀(jì)特別是近十幾年以來,礦床學(xué)研究正面臨著空前嚴(yán)峻的挑戰(zhàn)。隨著全球及中國社會發(fā)展進(jìn)入新的階段,囿于傳統(tǒng)科學(xué)技術(shù)的限制,發(fā)現(xiàn)新礦床的難度也與日俱增,保持礦產(chǎn)資源持續(xù)供應(yīng)問題正在逐年加劇,這些挑戰(zhàn)不但涵蓋了學(xué)科發(fā)展前沿,也涵蓋了國家戰(zhàn)略布局以及社會發(fā)展重大需求等諸多領(lǐng)域[9]。成礦模式創(chuàng)新研究不足、學(xué)科交叉融合程度不夠以及應(yīng)用基礎(chǔ)研究相對薄弱等方面是制約學(xué)科發(fā)展的主要方面。
成礦模式是對成礦作用的地質(zhì)條件、控礦要素、地質(zhì)過程、成礦機理和構(gòu)造背景的高度概括,既來源于找礦實踐,又指導(dǎo)找礦實踐[1]。
圖1 傳統(tǒng)成礦模式建立流程
成礦模式的建立工作在21世紀(jì)初達(dá)到高峰,部分學(xué)者對中國主要礦床類型的礦床模式(模型)也進(jìn)行了詳細(xì)的總結(jié)[1,22]。傳統(tǒng)礦床研究和成礦模式的建立過程雖然較多,但是概括來講一般包括野外數(shù)據(jù)采集→巖石樣品測試→數(shù)據(jù)分析→圖解制作→礦床成因類比→成礦模式建立等6個步驟(圖1)?;究梢詣澐譃閿?shù)據(jù)獲取和歸納演繹兩大階段:如野外描述、產(chǎn)狀測量、礦體編錄,乃至礦石分析測試、儲量計算等均屬于數(shù)據(jù)獲取階段;而對數(shù)據(jù)的歸納演繹、對比模擬、揭露成礦背景、構(gòu)建成礦模式、分析成礦系統(tǒng)、預(yù)測指導(dǎo)找礦等均屬于歸納演繹階段(表1)。兩個階段的工作手段的主觀性和客觀性不盡相同,有的工作手段兼有主觀與客觀兩種特性。如果把“主觀”、“主觀+客觀”以及“客觀”分別賦值為1、2、3,便可得到圖2的直觀表達(dá)。
表1 礦床學(xué)研究階段劃分
在成礦模式建立的過程中,尤其是礦床成因類比階段,主要依賴地質(zhì)工作者現(xiàn)有的知識水平來完成(圖1),所以傳統(tǒng)成礦模式的建立,雖同一礦床,因地質(zhì)工作者們的知識水平和對問題的理解角度不同而千差萬別,甚至互相矛盾。
圖 2 礦床研究方法主客觀因素直觀圖
近年來,有關(guān)新的成礦模式問世的理論創(chuàng)新報道相對減少[9],主要是圍繞已有成礦模式作進(jìn)一步的精細(xì)化工作(表2)。以研究程度較為成熟的斑巖型銅礦床為例,廣大學(xué)者對斑巖型銅礦的成因機制存有俯沖與碰撞兩種不同構(gòu)造背景的成礦體系解釋[23-24],尤其是對于成礦物質(zhì)來源、成礦過程、保存條件及宏觀控礦要素的確定等問題的研究還不夠成熟。眾所周知,加強對這些方面的深入研究必會有效推動斑巖銅礦成礦模式的創(chuàng)新進(jìn)程[9],但是基于傳統(tǒng)的思維方式和算法手段的限制,欲實現(xiàn)這一目標(biāo)尚有較大的阻力。
表 2 典型傳統(tǒng)成礦模式的研究特點
幾十年來,礦床學(xué)的研究總體上偏向?qū)Τ傻V理論,特別是成礦機制和模式創(chuàng)新的精細(xì)化基礎(chǔ)類研究,而缺乏將成礦機制與勘查指示緊密結(jié)合的應(yīng)用研究[9]。一直以來,國內(nèi)、外均十分重視礦產(chǎn)勘查手段的改進(jìn)和勘查效率的提升,但伴隨著更多的勘查技術(shù)和勘查方法的投入使用,勘查投入的成本也節(jié)節(jié)攀升。近年來我國地勘界也致力于“快速勘查方法體系”的研究開發(fā),但實際效果卻并不十分明顯[9]。
進(jìn)入21世紀(jì)后,各種先進(jìn)的實驗分析技術(shù)蓬勃發(fā)展,有力支撐了礦床成因的研究,雖然較為豐富了成礦理論系統(tǒng)創(chuàng)新成果,深刻揭示了地幔柱活動[42-44]、多塊體拼合[45]、克拉通破壞[46-49]、大陸裂谷作用[50-51]、板塊俯沖[52-53]、大陸碰撞[54-56]等重大地質(zhì)事件與大規(guī)模成礦作用的耦合關(guān)系,并在大陸動力學(xué)成礦[58]、大面積低溫成礦作用[59]等重大科學(xué)問題,產(chǎn)生了重要的國際影響[1],但是受傳統(tǒng)技術(shù)和思維方式的限制,研究者只能基于抽樣調(diào)查的手段建立各種類型的成礦模式,對成礦規(guī)律難以形成更深度的把握。
對于成礦模式的指導(dǎo)礦產(chǎn)勘查工作,首先應(yīng)加強對礦床研究的準(zhǔn)確化和精細(xì)化解剖,從而進(jìn)一步校正已有的成礦模式,并為發(fā)現(xiàn)和建立新的成礦模式提供可能。因此,要求研究者們能突破傳統(tǒng)思維禁錮,圍繞成礦機制多維探索,深刻揭露礦床成因的內(nèi)在規(guī)律,實現(xiàn)原有成礦模式的再創(chuàng)新,完善礦床學(xué)成礦理論研究的科學(xué)性和對礦床實體認(rèn)識的客觀性。
20世紀(jì)70年代,程裕淇、陳毓川、趙一鳴等在系統(tǒng)研究全國鐵礦成礦規(guī)律的基礎(chǔ)上,率先提出“鐵礦成礦系列”和“礦床成礦系列”的概念[60],使成礦系列成為之后20多年國內(nèi)礦床學(xué)研究的前沿課題[1]。對成礦系列和成礦體系的深入研究,有助于在宏觀視野揭露成礦規(guī)律,對指導(dǎo)礦產(chǎn)勘查工作有十分重要的意義。在研究成礦系列時間與空間結(jié)構(gòu)基礎(chǔ)上,發(fā)現(xiàn)了成礦系列的概況性、共同性、分帶性、重疊性、有序性、互補性和預(yù)見性等特點[61]。
與單個礦床的成因揭示不同,成礦體系與成礦規(guī)律的研究更需要應(yīng)用“大數(shù)據(jù)”理念,以“全息”式研究補充和完善傳統(tǒng)的“抽樣調(diào)查”的不足,以全新的視角揭露成礦系列和成礦規(guī)律[7]。
在傳統(tǒng)的成礦系列和成礦規(guī)律研究過程中,雖然廣大研究者們堅持秉持嚴(yán)密邏輯的思維過程,根據(jù)掌握的各類地質(zhì)資料,通過各種傳統(tǒng)的計算方法進(jìn)行建立并在指導(dǎo)找礦中進(jìn)行補充和完善,由于各人所掌握的數(shù)據(jù)類型有限,工作經(jīng)歷不同,對客觀問題的認(rèn)識千差萬別,且以往的工作中缺少一套規(guī)范性的技術(shù)要求[62],所以對成礦系列和成礦規(guī)律的歸納總結(jié)上存在視角單一、思維受限的弊端,在對數(shù)據(jù)利用和深度挖掘上存在一定的技術(shù)瓶頸,難免在成礦預(yù)測中做出主觀的估計和偏離實際的推斷。
“大數(shù)據(jù)”具有Volume(大量)、Velocity(高速)、Variety(多樣)和Value(價值)的特點[7-8],地質(zhì)數(shù)據(jù)便符合這“4V”特點。傳統(tǒng)的主流方法雖然也有數(shù)理統(tǒng)計學(xué)和數(shù)據(jù)庫管理等手段,但是在大數(shù)據(jù)特別是超算技術(shù)的應(yīng)用,大數(shù)據(jù)挖掘、機器學(xué)習(xí)技術(shù)逐漸成為眾多學(xué)者關(guān)注的熱點[63-64]。
隨著各種地球化學(xué)理論與方法在礦床學(xué)研究中的廣泛應(yīng)用,極大地提高了人們對礦床成因和成礦機制的認(rèn)識,不斷地豐富和完善成礦理論體系[9]。各類地質(zhì)數(shù)據(jù)的采集和挖掘都與信息社會的“大數(shù)據(jù)”不謀而合[7];礦床學(xué)研究經(jīng)過百余年的探索發(fā)展,尤其是20世紀(jì)70年代以來,元素地球化學(xué)[65-70]、同位素地球化學(xué)[71-74]、流體包裹體研究[75-78]、成礦年代學(xué)[79-81]、礦田構(gòu)造解析[82]以及成礦實驗方法[52, 83]等技術(shù)手段的廣泛應(yīng)用,使得礦床學(xué)研究方法不斷成熟和豐富,形成了龐大的數(shù)據(jù)集合。
當(dāng)前我國地質(zhì)學(xué)科已呈現(xiàn)數(shù)據(jù)密集型的大數(shù)據(jù)特點,而且每年仍以較高的速度產(chǎn)生,種類也不斷趨于完整豐富。網(wǎng)絡(luò)世界中通過各種模型所計算出來的數(shù)據(jù)比重也許正在被地質(zhì)工作實測的“實體”數(shù)據(jù)的比重所超越[7]。地學(xué)各個專業(yè)都在建立和規(guī)范管理各專業(yè)子領(lǐng)域的數(shù)據(jù)資源[7],雖然地質(zhì)原始數(shù)據(jù)已經(jīng)基本實現(xiàn)了海量積累,但是礦產(chǎn)資源綜合利用程度依然偏低,隨著地質(zhì)找礦難度日益增大,如果不利用現(xiàn)代“大數(shù)據(jù)”新技術(shù)對其進(jìn)行深度挖掘和關(guān)聯(lián)研究,將造成現(xiàn)有龐大數(shù)據(jù)資源的浪費。
地質(zhì)大數(shù)據(jù)具有“大數(shù)據(jù)”的大量性、高速性、多樣性和價值性的“4V”特征。
大量性:當(dāng)前的地質(zhì)數(shù)據(jù)是一個龐大的數(shù)據(jù)集合,包括地、礦、物、化、遙等各個領(lǐng)域[7]。2015年以來,中國地質(zhì)調(diào)查局系統(tǒng)不但建成區(qū)域地質(zhì)數(shù)據(jù)庫、基礎(chǔ)地質(zhì)數(shù)據(jù)庫、礦產(chǎn)資源數(shù)據(jù)庫、油氣能源數(shù)據(jù)庫,還建成了鉆探、遙感調(diào)查、地球物理、地球化學(xué)、水工環(huán)災(zāi)害、資料文獻(xiàn)和專題專項等十余個數(shù)據(jù)庫,且數(shù)據(jù)庫的種類以及數(shù)據(jù)都在不斷的充實與完善[7]。
高速性:由于新的探測技術(shù)或新的測試方法的引進(jìn)與運用,地質(zhì)數(shù)據(jù)的產(chǎn)生速度正在以成倍甚至幾何級數(shù)的速度累積增高,航空遙感、航空物探和區(qū)域地球化學(xué)等領(lǐng)域尤為顯著[7]。
多樣性:地質(zhì)數(shù)據(jù)從原始信號、原始數(shù)據(jù)到文字、圖像、音頻等結(jié)構(gòu)化與非結(jié)構(gòu)化數(shù)據(jù)類型無所不包。
價值性:大量的地質(zhì)數(shù)據(jù),尤其是礦產(chǎn)或物化探異常信息等數(shù)據(jù),具有較大的經(jīng)濟潛在價值。
除此之外,地質(zhì)數(shù)據(jù)還具有多元性、多維性、多源性、異構(gòu)性、時空性、方向性、相關(guān)性、隨機性、模糊性、時空不均勻性和過程的非線性等特點[84]。
近年來,人工智能各類算法已日趨成熟,深度學(xué)習(xí)(Deep Learning,DL)已經(jīng)在圖像分析[85]和語音識別[86]領(lǐng)域得到廣泛應(yīng)用,在自然語言處理[87]乃至視頻分類[88]等方面取得了令人矚目的成績。
由于機器學(xué)習(xí)的概念在生活中剛剛流行,所以很多人對專有名詞和概念尚較為模糊,對于人工智能、機器學(xué)習(xí)、深度學(xué)習(xí)等之間的從屬關(guān)系也不甚清楚。
機器學(xué)習(xí)(Machine Learning)是指用某些算法指導(dǎo)計算機利用已知數(shù)據(jù)得出適當(dāng)?shù)哪P?,并利用此模型對新的情境給出判斷的過程[64],通常分為有監(jiān)督學(xué)習(xí)(Supervised Learning)和無監(jiān)督學(xué)習(xí)(Unsupervised Learning)。有監(jiān)督學(xué)習(xí)又可分為分類(Classification)和回歸(Regression)兩類。其中分類算法包括:樸素貝葉斯算法、決策樹算法、Logistics回歸、臨近算法和支持向量機算法等;回歸算法包括:線性回歸和多項式回歸等。無監(jiān)督學(xué)習(xí)通??煞譃榫垲悾–lustering)、降維(Dimensionality Reduction)和關(guān)聯(lián)規(guī)則學(xué)習(xí)(Association rule learning)等算法。其中均值聚類、Mean-Shift、DBSCAN、支持向量機等屬于聚類算法范疇;主成分分析、奇異值分解、潛在狄里克雷特分配、潛在語義分析以及t-SEN等屬于降維算法范疇;Apriori、Euclat、EP-growth屬于關(guān)聯(lián)規(guī)則學(xué)習(xí)的范疇。神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)是機器學(xué)習(xí)的子集,深度學(xué)習(xí)(多層神經(jīng)網(wǎng)絡(luò))又是神經(jīng)網(wǎng)絡(luò)學(xué)習(xí)的子集(圖 3)。
學(xué)習(xí)的學(xué)習(xí)過程是一種對于有效特征的抽取過程(圖4)。深度學(xué)習(xí)的訓(xùn)練模型往往需要海量數(shù)據(jù)作為支撐[64]。深度神經(jīng)網(wǎng)絡(luò)可在不須輸入和輸出之間精確的數(shù)學(xué)表達(dá)式的情況下自動提取二者之間的映射關(guān)系[89],所以使用深度神經(jīng)網(wǎng)絡(luò)既能提取狀態(tài)特征,又易于訓(xùn)練[90]。
圖 3 人工智能算法關(guān)系示意圖
圖4 深度學(xué)習(xí)的學(xué)習(xí)過程示意圖(圖片引自網(wǎng)絡(luò):https://www.tooopen.com/view/179052.html和https://www.zhihu.com/question/264417928,有改動)
Mnih,等(2013)提出的深度Q學(xué)習(xí)網(wǎng)絡(luò)(Deep Q-network,DQN)算法,將卷積神經(jīng)網(wǎng)絡(luò)(Convolutional Neural Network,CNN)與強化學(xué)習(xí)(Reinforcement Learning,RL)中的經(jīng)典算法Q學(xué)習(xí)算法結(jié)合起來[90-91],是目前最為常用的一種深度學(xué)習(xí)算法。在基于價值的深度強化學(xué)習(xí)方法中,深度神經(jīng)網(wǎng)絡(luò)被用來當(dāng)作價值函數(shù)的逼近器,利用深度神經(jīng)網(wǎng)絡(luò)構(gòu)成策略網(wǎng)絡(luò)。將策略參數(shù)化表示擁有更好的收斂性,可以彌補在動作空間很大或動作為連續(xù)集的情況下DQN方法在解決機器人學(xué)問題時的局限性,更適用于高維連續(xù)空間的策略求解;多種結(jié)構(gòu)的深度強化學(xué)習(xí)算法可以使算法在計算效率、樣本利用率、計算資源上都有所提高[90]。如今卷積神經(jīng)網(wǎng)絡(luò)、深度Q學(xué)習(xí)網(wǎng)絡(luò)、多種結(jié)構(gòu)的深度強化學(xué)習(xí)、PILCO算法等,對人類社會的巨大影響已經(jīng)在多個行業(yè)有明顯的展現(xiàn)[92-97](表3)。
大數(shù)據(jù)挖掘特別適合于窺探具有多維性和全面性的現(xiàn)實世界,而高維數(shù)據(jù)處理是大數(shù)據(jù)深部挖掘的基礎(chǔ),它善于從支離破碎的信息中復(fù)原事物全貌[105]。圖5表示利用數(shù)據(jù)挖掘領(lǐng)域中拓?fù)鋽?shù)據(jù)分析(Topological Data Analysis,TDA)不僅可以有效地捕捉高維數(shù)據(jù)空間的拓?fù)湫畔?,還能夠在不丟失高維的信息的前提下有效降低大規(guī)模數(shù)據(jù)處理的維度[106]。
在最近幾年,隨著深地勘查需求的進(jìn)一步加強,以大數(shù)據(jù)分析和數(shù)值模擬為代表的新一輪科技革新正在不斷拓展著傳統(tǒng)學(xué)科的理論認(rèn)知領(lǐng)域[9]。利用人工智能深度學(xué)習(xí)技術(shù)對海量地質(zhì)數(shù)據(jù)進(jìn)行分析和計算,為深度探索隱藏在其背后的有價值的地質(zhì)信息資源提供了可能。
表 3 強化學(xué)習(xí)和深度強化學(xué)習(xí)部分算法列舉表
圖5 拓?fù)鋽?shù)據(jù)分析示例(據(jù)參考文獻(xiàn)[106])
大數(shù)據(jù)分析和機器學(xué)習(xí)等正逐步被應(yīng)用到地學(xué)領(lǐng)域中來[107],并取得一些具有啟發(fā)性的理論創(chuàng)新成果[108],如應(yīng)用關(guān)聯(lián)規(guī)則算法對與金礦相關(guān)的侵入巖、火山巖、變質(zhì)巖建造及區(qū)域構(gòu)造地質(zhì)大數(shù)據(jù)的關(guān)聯(lián)性的探索發(fā)現(xiàn)地質(zhì)要素之間的共生關(guān)系[109];利用貝葉斯網(wǎng)絡(luò)(Bayesian Network)揭示礦床的成因機制并構(gòu)建大數(shù)據(jù)-智能礦床成礦與找礦模型等[110];利用大數(shù)據(jù)手段對特提斯斑巖成礦帶、中亞斑巖成礦帶和環(huán)太平洋斑巖成礦帶進(jìn)行地質(zhì)地球化學(xué)對比分析,進(jìn)而揭示地殼性質(zhì)對斑巖銅礦控制作用等[111]。
最近,我國的很多學(xué)者正嘗試?yán)脵C器學(xué)習(xí)方法對地質(zhì)大數(shù)據(jù)進(jìn)行深度挖掘,陳進(jìn)等(2020)通過隨機森林算法對大尹格莊金礦的三維礦體定位預(yù)測,圈定了7個三維找礦靶區(qū),取得了較好的效果[112];劉艷鵬等(2020)如通過卷積神經(jīng)網(wǎng)絡(luò)方法挖掘元素Zn在地表的分布特征與礦體在地下空間就位的耦合關(guān)系,得到了準(zhǔn)確率為95%的CNN模型[113];閆巖等(2021)利用數(shù)據(jù)信息搭建變質(zhì)程度-礦物網(wǎng)絡(luò)模型,揭露綠片巖相和麻粒巖相對金屬礦床和非金屬礦床的控制關(guān)系,較好的證明了數(shù)據(jù)挖掘方法有效性[114];張振杰等(2021)利用決策樹算法、神經(jīng)網(wǎng)絡(luò)、樸素貝葉斯和支持向量機等算法對閩西南馬坑式鐵礦進(jìn)行成礦預(yù)測,四種方法均取得了較好的預(yù)測效果[115]。此外,基于多主題數(shù)據(jù)集成開展礦床三維可視化建模[116]、智能技術(shù)集成算法對非線性地球化學(xué)特征進(jìn)行精細(xì)化信息識別與預(yù)測預(yù)判[117]、隨機森林算法對磁鐵礦主微量元素在礦床成因分類中的重要性排序[118]以及利用成本敏感神經(jīng)網(wǎng)絡(luò)預(yù)測鐵多金屬遠(yuǎn)景[119]等研究也相繼報道。
大數(shù)據(jù)時代給礦床學(xué)研究的發(fā)展帶來了新的啟迪,但是在實際操作過程中也面臨著許多亟待解決的科學(xué)問題。如在礦床研究中利用地物化遙探測技術(shù)獲取各種原始數(shù)據(jù)具有結(jié)構(gòu)化、半結(jié)構(gòu)化和非結(jié)構(gòu)化特征[120],如此對地質(zhì)數(shù)據(jù)的儲存管理技術(shù)則提出了更高的要求。又如地質(zhì)體、地質(zhì)結(jié)構(gòu)和地質(zhì)過程的復(fù)雜性、不可見性和數(shù)據(jù)采集的抽樣方式,導(dǎo)致出現(xiàn)“結(jié)構(gòu)”“關(guān)系”“參數(shù)”“演化”等信息不全的狀況,需要對地質(zhì)數(shù)據(jù)進(jìn)行三維、動態(tài)、多尺度、多細(xì)節(jié)層次的可視化建模技術(shù)提出更高的要求[120-121]。再如礦床大數(shù)據(jù)的多參數(shù)、多時態(tài)、多模態(tài)、多尺度以及不確定性特征,對基于全體數(shù)據(jù)的數(shù)據(jù)挖掘技術(shù)的運用還需要進(jìn)一步的嘗試。
目前,人工智能在礦床學(xué)中的相關(guān)研究還相對較少。對于利用人工智能進(jìn)行礦床大數(shù)據(jù)深部挖掘,成礦模式建立和高效勘查的研究工作還處于起步階段,當(dāng)前尚未利用大數(shù)據(jù)手段對不同種類的礦床類型特征開展廣泛研究,而對用于進(jìn)行機器學(xué)習(xí)的數(shù)據(jù)平臺也未能進(jìn)行嚴(yán)格篩選,因而導(dǎo)致部分大數(shù)據(jù)手段和人工智能方法在地質(zhì)學(xué)研究中所取得的效果不佳也在所難免,要實現(xiàn)“智能高效勘查”的要求還需繼續(xù)努力。
構(gòu)建地球科學(xué)大數(shù)據(jù)挖掘與機器學(xué)習(xí)的基本框架,運用機器學(xué)習(xí)關(guān)聯(lián)規(guī)則算法[84]、高維數(shù)據(jù)降維[122-125]、推薦系統(tǒng)算法[64]、大圖形社區(qū)結(jié)構(gòu)識別[105]、人工智能地質(zhì)學(xué)的建模[107]、分類與預(yù)測[126]及流數(shù)據(jù)處理[127-128]等技術(shù),對已有礦產(chǎn)勘查方法與成礦模式進(jìn)行深度挖掘并提煉勘查標(biāo)識體系,從技術(shù)上彌補“無法用常規(guī)軟件進(jìn)行數(shù)據(jù)管理”[7]的不足已成為可能。
1)進(jìn)入21世紀(jì)特別是近十幾年以來,礦床學(xué)研究正面臨著空前嚴(yán)峻的挑戰(zhàn),囿于傳統(tǒng)科學(xué)技術(shù)的限制,發(fā)現(xiàn)新礦床的難度也與日俱增。成礦模式創(chuàng)新研究不足、學(xué)科交叉融合程度不夠以及應(yīng)用基礎(chǔ)研究相對薄弱等方面是制約學(xué)科發(fā)展的主要方面。
2)傳統(tǒng)礦床研究和成礦模式的建立一般包括野外數(shù)據(jù)采集→巖石樣品測試→數(shù)據(jù)分析→圖解制作→礦床成因類比→成礦模式建立等6個步驟,基本可以劃分為數(shù)據(jù)獲取和歸納演繹兩大階段。在傳統(tǒng)成礦模式建立的過程中,尤其是礦床成因類比階段,主要依賴地質(zhì)工作者現(xiàn)有的知識水平來完成,雖同一礦床,因研究者們的知識水平和對問題的理解角度不同而千差萬別,甚至互相矛盾。
3)當(dāng)前地學(xué)各個專業(yè)都在建立和規(guī)范管理各專業(yè)子領(lǐng)域的數(shù)據(jù)資源,地質(zhì)學(xué)科已呈現(xiàn)數(shù)據(jù)密集型的大數(shù)據(jù)(4V)特點,而且每年仍以較高的速度產(chǎn)生,但是礦產(chǎn)資源綜合利用程度依然偏低。
4)機器學(xué)習(xí)通常分為有監(jiān)督學(xué)習(xí)和無監(jiān)督學(xué)習(xí)。有監(jiān)督學(xué)習(xí)又可分為分類和回歸兩類。無監(jiān)督學(xué)習(xí)通??煞譃榫垲?、降維和關(guān)聯(lián)規(guī)則學(xué)習(xí)等算法。利用人工智能深度學(xué)習(xí)技術(shù)對海量地質(zhì)數(shù)據(jù)進(jìn)行分析和計算,為深度探索隱藏在其背后的有價值的地質(zhì)信息資源提供了可能。
5)當(dāng)前,我國部分學(xué)者通過機器學(xué)習(xí)方法在地質(zhì)大數(shù)據(jù)深度挖掘方面的研究工作已經(jīng)開展了嘗試,并取得了一定的成效。構(gòu)建地球科學(xué)大數(shù)據(jù)挖掘與機器學(xué)習(xí)的基本框架,運用機器學(xué)習(xí)關(guān)聯(lián)規(guī)則算法、高維數(shù)據(jù)降維、推薦系統(tǒng)算法、大圖形社區(qū)結(jié)構(gòu)識別、人工智能地質(zhì)學(xué)的建模、分類與預(yù)測及流數(shù)據(jù)處理等技術(shù),對已有礦產(chǎn)勘查方法與成礦模式進(jìn)行深度挖掘并提煉勘查標(biāo)識體系,從技術(shù)上彌補“無法用常規(guī)軟件進(jìn)行數(shù)據(jù)管理”的不足已成為可能。
[1] 李建威,趙新福,鄧曉東,等.新中國成立以來中國礦床學(xué)研究若干重要進(jìn)展[J].中國科學(xué):地球科學(xué),2019,49(11):1720-1771.
LI Jianwei, ZHAO Xinfu, DENG Xiaodong, et al. An overview of the advance on the study of China’s ore deposits during the last seventy years[J]. Scientia Sinica Terrae, 2019, 49 (11): 1720–1771.
[2] McCulloch Warren S, Pitts Walter. A Logical Calculus of the Ideas Immanent in Nervous Activity[J]. Bulletin of Mathematical Biophysics, 1943 (5): 115-133.
[3] Hebb D O. The Organization of Behavior[M]. New York: Wiley, 1949.
[4] Rosenblatt F. The perceptron: A probabilistic model for information storage and orgnization in the brain[J]. Psychological Review, 1958 (65): 388-408.
[5] Minsky M, Papert S. Perceptrons[M]. Cambridge: MIT Press, 1969.
[6] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagaing errors[J]. Nature, 1986 (323): 533-536.
[7] 王登紅,劉新星,劉麗君.地質(zhì)大數(shù)據(jù)的特點及其在成礦規(guī)律、成礦系列研究中的應(yīng)[J].礦床地質(zhì),2015,34(6):1143-1154.
WANG Denghong, LIU Xinxing, LIU Lijun. Characteristics of Big Geodata and its application to study of minerogenetic regularity and minerogenetic series[J]. Mineral Deposits, 2015, 34 (6): 1143-1154.
[8] Mayer-Sch?nhberger V, Cukier K. Big Data: A revolution that will transform how we live, work and think[M]. New York: Houghton Mifflin Harcourt Publishing Company, 2013, 1-158.
[9] 陳華勇.對我國礦床學(xué)未來發(fā)展方向的思考[J].地學(xué)前緣,2020,27(2):99-105.
CHEN Huayong. Meditations on the future development of ore deposit science in China[J]. Earth Science Frontiers, 2020, 27 (2): 99-105.
[10] 羅雪,曹新志.礦床學(xué)研究思路的變化與進(jìn)展綜述[J]. 地質(zhì)找礦論叢,2010,25(2):147-152.
LUO Xue, CAO Xinzhi. Review on the change and development of the research thoughts about mineral deposit geology[J]. Contributions to Geology and Mineral Resources Research, 2010, 25 (2): 147-152.
[11] 袁見齊,朱上慶,翟裕生.礦床學(xué)[M].北京:地質(zhì)出版社,1985: 1-346.
YUAN Jianqi, ZHU Shangqing, ZHAI YuSheng. Mineral Deposits[M]. Beijing: Geological Publishing House, 1985: 1-346.
[12] 翟裕生.關(guān)于礦床學(xué)研究前景的探討[J].礦床地質(zhì),1999,18(2):146-152.
ZHAI Yusheng. A discussion on the geology of ore deposits[J]. Mineral Deposits, 1999a, 18 (2): 146-152 .
[13] 薛春紀(jì),祁思敬,隗合明.基礎(chǔ)礦床學(xué)[M].北京:地質(zhì)出版社,2006: 1-27.
XUE Chunji, QI Sijing, WEI Heming. Essentials of Mineral Deposits[M]. Beijing: Geological Publishing House, 2006: 1-27.
[14] Grilhaumou N. Characterization of hydrocarbon fluid inclusions by infrared fluorescence microscopetrometry[J]. Mineralogical Magazine, 1990, 54: 519-533.
[15] 關(guān)廣岳.成礦動力系統(tǒng)的定態(tài)與失定[J].東北大學(xué)學(xué)報,1994,15(1):20-23.
GUAN Guangyue. Steady and unsteady in dynamic systems of mineralization[J]. Journal of Northeastern University, 1994, 15(1): 20-23.
[16] 李人澍.成礦系統(tǒng)分析的理論與實踐[M].北京:地質(zhì)出版社,1996: 1-262.
LI Renshu. Theory and practice of metallogenic system analysis[M]. Beijing: Geological Publishing House, 1996: 1-262.
[17] 李人澍,朱華平.成礦系統(tǒng)的結(jié)構(gòu)與聚礦功能[J].地學(xué)前緣,1999,6(1):103-113.
LI Renshu, ZHU Huaping. System structure and ore-forming function[J]. Earth Science Frontiers, 1999, 6(1): 103-113.
[18] 侯增謙,李紅陽.試論幔柱構(gòu)造與成礦系統(tǒng):以三江特提斯成礦域為例[J].礦床地質(zhì),1998,17(2):9-113.
HOU Zengqian, LI Hongyang. A tentative discussion on the mantle plume tectonics and metallogenic system as exemplified by the Sanjiang tethyan metallogenic domain[J]. Mineral Deposits, 1998, 17 (2): 9-113.
[19] 翟裕生.論成礦系統(tǒng)[J].地學(xué)前緣,1999,6(1):12-27.
ZHAI Yusheng. On the metallogenic system [J]. Earth Science Frontiers, 1999, 6(1): 12-27.
[20] 翟裕生.成礦系統(tǒng)研究與找礦[J].地質(zhì)調(diào)查與研究,2003,26(2):65-71.
ZHAI Yusheng. Research on metallogenic system[J]. Geological Survey and Research, 2003, 26(2): 65-71.
[21] 范永香,陽正熙.成礦規(guī)律與成礦預(yù)測[M]. 徐州:中國礦業(yè)大學(xué)出版社,2005: 1-274.
FAN Yongxiang, YANG Zhengxi. Metallogenetic regularities and prediction[M].Xuzhou: China University of Mining and Technology Press, 2005: 1-274.
[22] 毛景文,張作衡,裴榮富. 中國礦床模型概論[M]. 北京:地質(zhì)出版社,2012: 1-560.
MAO Jingwen, ZHANG Zuoheng, PEI Rongfu.Mineral deposit models in China[M]. Beijing: Geological Publishing House, 2012: 1-560.
[23] Sillitoe R H. Porphyry copper systems[J]. Economic Geology, 2010, 105 (1): 3-41.
[24] 侯增謙,鄭遠(yuǎn)川,楊志明,等. 大陸碰撞成礦作用:I. 岡底斯新生代斑巖成礦系統(tǒng)[J]. 礦床地質(zhì),2012,(4):647-670.
HOU Zengqian, ZHENG Yuanchuan, YANG Zhiming, et al. Metallogenesis of continental collision setting: Part I. Gangdese Cenozoic porphyry Cu-Mo systems in Tibet[J]. Mineral Deposits, 2012, 31(4): 647-670.
[25] 周琦,杜遠(yuǎn)生,覃英. 古天然氣滲漏沉積型錳礦床成礦系統(tǒng)與成礦模式:以黔湘渝毗鄰區(qū)南華紀(jì)“大塘坡式”錳礦為例[J]. 礦床地質(zhì),2013,32:457-466.
ZHOU Qi, DU Yuansheng, TAN Ying. Ancient natural gas seepage sedimentary-type manganese metallogenic system and ore-forming model: a case study of Datangpo type manganese deposits formed in rift basin of Nanhua Period along Guizhou-Hunan-Chongqing border area [J]. Mineral Deposits, 2013, 32: 457-466.
[26] 周琦,杜遠(yuǎn)生,袁良軍,等. 古天然氣滲漏沉積型錳礦床找礦模型:以黔湘渝毗鄰區(qū)南華紀(jì)“大塘坡式”錳礦為例[J]. 地質(zhì)學(xué)報,2017,91:2285-2298.
ZHOU Qi, DU Yuansheng, YUAN Liangjun, et al. Exploration models of ancient natural gas seep sedimentary-type manganese ore deposit: a case study of the Nanhua period “Datangpo” type manganese ore in the conjunction area of Guizhou, Hunan and Chongqing[J]. Acta Geologica Sinica, 2017, 91: 2285-2298.
[27] ZHAO C, Ni P, Wang G G, et al. Geology, fluid inclusion, and isotope constraints on ore genesis of the Neoproterozoic Jinshan orogenic gold deposit, South China[J]. Geofluids, 2013, 13 (4): 506-527.
[28] 毛光周,華仁民,龍光明,等.江西金山金礦成礦時代探討:來自石英流體包裹體Rb-Sr 年齡的證據(jù).地質(zhì)學(xué)報,2008,82(4):532-539.
MAO Guangzhou, HUA Renmin, LONG Guangming, et al. Discussion on the mineralogenetic epoch of the Jinshan gold deposit, Jiangxi Province: Based on the quartz fluids inclusion Rb-Sr dating. Acta Geologica Sinica, 2008, 82 (4): 532-539.
[29] 張金春.江西金山韌性剪切型金礦成礦地球化學(xué)研究[D].南京:南京大學(xué),1994.
ZHANG Jinchun. Metallogenic and geochemistry of Jinshan ductile sheartype gold deposit, Jiangxi Province[D]. Nanjing: Nanjing University, 1994.
[30] 黃宏立,楊文思.贛東北金山金礦床的地質(zhì)特征及礦床成因[J].地質(zhì)找礦論叢,1990,5(2):29-39.
HUANG Hongli, YANG Wensi. Geological characteristics and genesis of Jinshan gold deposit in the northeastern Jiangxi Province[J]. Contributions to Geology and Mineral Resources Research, 1990, 5 (2): 29-39.
[31] XU D R, Deng T, Chi G X, et al. Gold mineralization in the Jiangnan Orogenic Belt of South China: Geological, geochemical and geochronological characteristics, ore deposit-type and geodynamic setting[J]. Ore Geology Reviews, 2017, 88: 565-618.
[32] WANG C Y, Zhou M F, Qi L. Chalcophile element geochemistry and petrogenesis of high-Ti and low-Ti magmas in the Permian Emeishan large igneous province, SW China[J]. Contrib Mineral Petrol, 2011, 161: 237-254.
[33] WANG Q, Xu J F, Jian P, et al. Petrogenesis of adakitic porphyries in an extensional tectonic setting, Dexing, South China: Implications for the genesis of porphyry copper mineralization[J]. Journal of Petrology, 2006, 47(1): 119-144.
[34] 王國光,倪培,趙超,等.德興大型銅金礦集區(qū)的研究進(jìn)展和成礦模式[J].巖石學(xué)報,2019,35(12):3644-3658.
WANG Guoguang, NI Pei, ZHAO Chao, et al. The research advances and genetic model of the giant Dexing Cu-Au ore cluster[J]. Acta Petrologica Sinica, 2019, 35 (12): 3644-3658
[35] 翟裕生,姚書振,林新多,等.長江中下游地區(qū)鐵銅(金)成礦規(guī)律[M].北京:地質(zhì)出版社.1992: 1-235.
ZHAI Yusheng, YAO Shuzhen, LIN Xinduo, et al. Regularities of Metallogenesis for Copper (Gold) Deposits in the Middle and Lower Reaches of the Yangtze River Area[M]. Beijing: Geological Publishing House, 1992: 1-235.
[36] 黃許陳,儲國正.銅陵獅子山礦田多位一體(多層樓)模式[J].礦床地質(zhì),1993,3:221-230.
HUANG Xuchen, CHU Guozheng. Multistory metallogenic model of the Shizishan orefield in Tongling, Anhui Province[J]. Mineral Deposits, 1993, 3: 221-230.
[37] ZHAO X F, Zhou M F, Gao J F, et al. In situ Sr isotope analysis of apatite by LA-MC-ICPMS: Constraints on the evolution of ore fluids of the Yinachang Fe-Cu-REE deposit, Southwest China[J]. Miner Depos, 2015, 50: 871-884.
[38] LI X C, Zhao X F, Zhou M F, et al. Fluid inclusion and isotopic constraints on the origin of the Paleoproterozoic Yinachang Fe-Cu-(REE) Deposit, Southwest China[J]. Econ Geol, 2015, 110: 1339-1369.
[39] SU Z K, Zhao X F, Li X C, et al. Using elemental and boron isotopic compositions of tourmaline to trace fluid evolutions of IOCG systems: The worldclass Dahongshan Fe-Cu deposit in SW China[J]. Chem Geol, 2016, 441: 265-279.
[40] ZHAO X F, Zhou M F, Su Z K, et al. Geology, geochronology, and geochemistry of the Dahongshan Fe-Cu-(Au-Ag) deposit, Southwest China: Implications for the formation of iron oxide copper-gold deposits in intracratonic rift settings[J]. Econ Geol, 2017, 112: 603-628.
[41] 康永孚,苗樹屏,李崇佑,等.中國礦床編委會編《中國礦床》中冊:中國鎢礦床[M].北京:地質(zhì)出版社,1994: 1-104.
KANG Yongfu, MIAO Shuping, LI Chongyou, et al. China Mineral Deposits Editorial Committee, China Mineral Deposits, Volume Ⅱ (Tungsten ore deposits in China)[M]. Beijing: Geological Publishing House, 1994: 1-104.
[42] Cao J, Wang C Y, Xu Y G, et al. Triggers on sulfide saturation in Fe-Ti oxide-bearing, maficultramafic layered intrusions in the Tarim large igneous province, NW China[J]. Miner Depos, 2017, 52: 471-494.
[43] Wang C Y, Wei B, Zhou M F, et al. A synthesis of magmatic Ni-Cu-(PGE) sulfide deposits in the ~260 Ma Emeishan large igneous province, SW China and northern Vietnam[J]. Journal of Asian Earth Sciences, 2018, 154: 162-186.
[44] Zhang D Y, Zhang Z C, Huang H, et al. Petrogenesis and metallogenesis of the Wajilitag and Puchang Fe-Tioxide-rich intrusive complexes, orthwestern Tarim Large Igneous Province[J]. Lithos, 2018, 304-307: 412-435.
[45] 秦克章,翟明國,李光明,等.中國陸殼演化、多塊體拼合造山與特色成礦的關(guān)系[J].巖石學(xué)報,2017,33(2):305-325.
QIN Kezhang, ZHAI Mingguo, LI Guangming, et al. Links of collage orogenesis of multiblocks and crust evolution to characteristic metallogeneses in China[J]. Acta Petrologica Sinica, 2017, 33 (2): 305-325.
[46] 翟明國,范宏瑞,楊金輝,等.非造山帶型金礦:膠東型金礦的陸內(nèi)成礦作用[J].地學(xué)前緣,2004,11:85-98.
ZHAI Mingguo, FAN Hongrui, YANG Jinhui, et al. Large-scale cluster of gold deposits in east Shandong: anorogenic metallogenesis[J]. Earth Science Frontiers, 2004, 11: 85-98.
[47] 翟明國. 華北克拉通的形成演化與成礦作用[J].礦床地質(zhì),2010,29:24-33.
ZHAI Mingguo. Tectonic evolution and metallogenesis of North China Craton[J]. Mineral Deposits, 2010,29:24-33.
[48] 翟明國.華北前寒武紀(jì)成礦系統(tǒng)與重大地質(zhì)事件的聯(lián)系[J].巖石學(xué)報,2013,29:1759-1773.
ZHAI Mingguo. Secular changes of metallogenic systems link with continental evolving of the North China Craton[J]. Acta Petrologica Sinica, 2013,29 (5):1759-1773.
[49] Li J W,Bi S J,Selby D,et al. Giant Mesozoic gold provinces related to the destruction of the North China craton[J]. Earth Planet Sci Lett,2012,349-350: 26-37.
[50] Zhao X F,Zhou M F. Fe-Cu deposits in the Kangdian region, SW China: A Proterozoic IOCG (iron-oxide-copper-gold) metallogenic province[J]. Miner Depos, 2011, 46: 731-747.
[51] Zhou M F,Zhao X F,Chen W T,et al. Proterozoic Fe-Cu metallogeny and supercontinental cycles of the southwestern Yangtze Block, southern China and northern Vietnam[J]. Earth-Sci Rev, 2014, 139: 59-82.
[52] Zhang C Q, Wu Y, Hou L, et al. Geodynamic setting of mineralization of Mississippi Valley-type deposits in world-class Sichuan-Yunnan-Guizhou Zn-Pb triangle, southwest China: Impli- cations from age-dating studies in the past decade and the Sm-Nd age of Jinshachang deposit[J]. Journal of Asian Earth Sciences, 2015, 103: 103-114.
[53] Hu R Z, Fu S L, Huang Y, et al. The giant South China Mesozoic low temperature metallogenic domain: Reviews and a new geodynamic model[J]. Journal of Asian Earth Sciences, 2017, 137: 9-34.
[54] Hou Z Q, Gao Y F, Qu X M, et al. Origin of adakitic intrusives generated during mid-Miocene east-west extension in southern Tibet[J]. Earth Planet Sci Lett, 2004, 220: 139-155.
[55] Hou Z Q, Yang Z M, Qu X M, et al. The Miocene Gangdese porphyry copper belt generated during post-collisional extension in the Tibetan Orogen[J]. Ore Geol Rev, 2009, 36: 25-51.
[56] Hou Z Q, Zhou Y, Wang R, et al. Recycling of metal-fertilized lower continental crust: Origin of non-arc Au-rich porphyry deposits at cratonic edges[J]. Geology, 2017, 45: 563-566.
[57] Yang Z M, Hou Z Q, White N C, et al. Geology of the post-collisional porphyry copper-molybdenum deposit at Qulong, Tibet[J]. Ore Geol Rev, 2009, 36: 133-159.
[58] 胡瑞忠,毛景文,范蔚茗,等.華南陸塊陸內(nèi)成礦作用的一些科學(xué)問題[J].地學(xué)前緣,2010,17(2):13-26.
HU Ruizhong, MAO Jingwen, FAN Weiming, et al. Some scientific questions on the intra-continental metallogeny in the South China continent[J]. Earth Science Frontiers, 2010, 17(2): 13-26.
[59] 胡瑞忠,付山嶺,肖加飛.華南大規(guī)模低溫成礦的主要科學(xué)問題[J].巖石學(xué)報,2016,32(11):3239-3251.
HU Ruizhong, FU Shanling, XIAO Jiafei. Major scientific problems on low-temperature metallogenesis in South China[J]. Acta Petrologica Sinica, 2016, 32 (11): 3239-3251.
[60] 程裕淇,陳毓川,趙一鳴,等.再論礦床的成礦系列問題:兼論中生代某些礦床的成礦系列[J].地質(zhì)論評,1983,29:127-139.
CHENG Yuqi, CHEN Yuchuan, ZHAO Yiming, et al. Further discussion on the problems of minerogenetic series of mineral deposits[J]. Geological Review, 1983, 29: 127-139.
[61] 王明志,李閆華,鄢云飛,等.若干成礦預(yù)測理論研究綜述[J].資源環(huán)境與工程,2007,21(4):363-369.
WANG Mingzhi, LI Yanhua, YAN Yunfei, et al. Overview of research for Metallogenic forecastion[J]. Resources Environment & Engineering, 2007, 21 (4): 363-369.
[62] 王登紅,徐志剛,盛繼福,等.全國重要礦產(chǎn)和區(qū)域成礦規(guī)律研究進(jìn)展綜述[J].地質(zhì)學(xué)報,2014,88(12):2176-2191.
WANG Denghong, XU Zhigang, SHENG Jifu, et al. Progress on the study of regularity of major mineral Resources and regional metallogenic regularity in China: a review[J]. Acta Geologica Sinica, 2014, 88 (12): 2176-2191.
[63] 張旗,周永章.大數(shù)據(jù)正在引發(fā)地球科學(xué)領(lǐng)域一場深刻的革命—《地質(zhì)科學(xué)》2017年大數(shù)據(jù)專題代序[J].地質(zhì)科學(xué),2017,52(3):637-648.
ZHANG Qi, ZHOU Yongzhang. Big data will lead to a profound revolution in the field of geological science[J]. Chinese Journal of Geology, 2017, 52 (3): 637-648.
[64] 周永章,張良均,張奧多,等.地球科學(xué)大數(shù)據(jù)挖掘與機器學(xué)習(xí)[M].廣州:中山大學(xué)出版社,2018: 1-269.
ZHOU Yongzhang, ZHANG Liangjun, ZHANG Aoduo, et al. Big data mining & machine learning in geoscience[M]. Guangzhou: Sux Yat-sen University Press, 2018: 1-269.
[65] 張榮華,劉隆隆,陸成慶.長江中下游某些火山巖鐵礦微量元素的地球化學(xué)[J].中國地質(zhì)科學(xué)院院報,1979,1:90-110.
ZHANG Ronghua, LIU Longlong, LU Chengqing. On geochemistry of trace elements of several porphyrite iron deposits along the middle lower Changjiang valley[J]. Bulletin of the Chinese Academy of Geological Sciences, 1979, 1: 90-100.
[66] 林師整.磁鐵礦礦物化學(xué)、成因及演化的探討[J].礦物學(xué)報,1982,3:166-174.
LIN Shizheng. A contribution to the chemistry, origin and evolution of magnetite[J]. Acta Mineralogica Sinica, 1982, 3: 166-174.
[67] Song X Y, Keays R R, Zhou M F, et al. Siderophile and chalcophile elemental constraints on the origin of the Jinchuan Ni-Cu-(PGE) sulfide deposit, NW China[J]. Geochim Cosmochim Acta, 2009, 73: 404-424.
[68] 趙振華.微量元素地球化學(xué)原理[M].第2版. 北京:科學(xué)出版社,2016: 285-289.
ZHAO Zhenhua. Geochemistry of Trace Elements[M]. 2nd Edition . Beijing: Science Press, 2016: 285-289.
[69] Yan J, Hu R Z, Liu S, et al. Nano SIMS element mapping and sulfur isotope analysis of Au-bearing pyrite from Lannigou Carlin-type Au deposit in SW China: New insights into the origin and evolution of Au-bearing fluids[J]. Ore Geol Rev, 2018, 92: 29-41.
[70] Wu Y F, Fougerouse D, Evans K, et al. Gold, arsenic, and copper zoning in pyrite: A record of fluid chemistry and growth kinetics[J]. Geology, 2019, 47: 641-644.
[71] Chen L, Li X H, Li J W, et al. Extreme variation of sulfur isotopic compositions in pyrite from the Qiuling sediment-hosted gold deposit, West Qinling orogen, central China: An in situ SIMS study with implications for the source of sulfur[J]. Miner Depos, 2015, 50: 643-656.
[72] Yang D, Hou Z Q, Zhao Y, et al. Lithium isotope traces magmatic fluid in a seafloor hydrothermal system[J]. Sci Rep, 2015, 5: 13812.
[73] Fu J L, Hu Z Z, Zhang W, et al. In situ sulfur isotopes (δ34S and δ33S) analyses in sulfides and elemental sulfur using high sensitivity cones combined with the addition of nitrogen by laser ablation MC-ICP-MS[J]. Anal Chim Acta, 2016, 911: 14-26.
[74] Zheng Y C, Liu S A, Wu C D, et al. Cu isotopes reveal initial Cu enrichment in sources of giant porphyry deposits in a collisional setting[J]. Geology, 2018, 47: 135-138.
[75] Hou Z Q, Zaw K, Qu X M, et al. Origin of the Gacun volcanic-hosted massive sulfide deposit in Sichuan, China: Fluid inclusion and oxygen isotope evidence[J]. Econ Geol, 2001, 96: 1491-1512.
[76] Fan H R, Zhai M G, Yang K F, et al. Late Mesozoic gold mineralization in the North China Craton. In: Zhai M G, Zhao Y, Zhao T P, eds. Main Tectonic Events and Metallogeny of the North China Craton[J]. Singapore: Springer. 2016, 511-525.
[77] Chen P W, Zeng Q D, Zhou T C, et al. Evolution of fluids in the Dasuji porphyry Mo deposit on the northern margin of the North China Craton: Constraints from Microthermometric and LA-ICP-MS analyses of fluid inclusions[J]. Ore Geol Rev, 2019, 104: 26-45.
[78] Pan J Y, Ni P, Wang R C. Comparison of fluid processes in coexisting wolframite and quartz from a giant vein-type tungsten deposit, South China: Insights from detailed petrography and LAICP-MS analysis of fluid inclusions[J]. Am Miner, 2019, 104: 1092-1116.
[79] Qiu H N.40Ar/39Ar dating of the quartz samples from two mineral deposits in western Yunnan (SW China) by crushing in vacuum[J]. Chem Geol, 1996, 127: 211-222.
[80] Deng X D, Li J W, Luo T, et al. Dating magmatic and hydrothermal processes using andradite-rich garnet U-Pb geochronometry[J]. Contrib Mineral Petrol, 2017, 172: 71.
[81] Zhou Y T, Lai Y, Shu Q H, et al. Geochronology and fluid inclusion study of the Shabutai porphyry Mo deposit, Inner Mongolia[J]. Ore Geol Rev, 2017, 81: 745-759.
[82] 翟裕生,呂古賢.構(gòu)造動力體制轉(zhuǎn)換與成礦作用[J].地球?qū)W報,2002,23(2):97-102.
ZHAI Yusheng, LYU Guxian. Transition of Tectonic and Dynamic Regime and Mineralization[J]. Acta Geoscientia Sinica, 2002, 23 (2): 97-102.
[83] Liu X D, Lu X C, Wang R C, et al. Speciation of gold in hydrosulphide-rich ore-forming fluids: Insights from firstprinciples molecular dynamics simulations[J]. Geochim Cosmochim Acta, 2011, 75: 185-194.
[84] 翟明國,楊樹鋒,陳寧華,等.大數(shù)據(jù)時代:地質(zhì)學(xué)的挑戰(zhàn)與機遇[J].學(xué)科與領(lǐng)域,2018,825-831.
ZHAI Mingguo, YANG Shufeng, CHEN Ninghua, et al. Big data epoch: challenges and opportunities for geology[J]. Subject and Field, 2018, 825-831.
[85] Krizhevsky A, Sutskever I, Hinton G E. Imagenet classification with deep convolutional neural networks// Proceedings of the 26thAnnual Conference on Neural Information Processing Systems[J]. Nevada, USA, 2012, 1097-1105.
[86] Graves A, Mohamed A, Hinton G. Speech recognition with deep recurrent neural networks// Proceedings of the IEEE Conferenc on Acoustics[J]. Speech and Signal Processing Vancouver, Canada, 2013, 6645-6649.
[87] Cho K, Merrienboer B V, Gulcehre C, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation// Proceeding of the Conference on Empirical Methods in Natural Language Processing[J]. Doha, Qatar, 2014, 1724-1734.
[88] Karpathy A, Toderici George, Shetty S, et al. Large-scale video classification with convolutional neural networks// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognnition[J]. Columbus, USA, 2014, 1725-1732.
[89] 劉艷鵬,朱立新,周永章.卷積神經(jīng)網(wǎng)絡(luò)及其在礦床找礦預(yù)測中的應(yīng)用研究:以安徽省兆吉口鉛鋅礦床為例[J].巖石學(xué)報,2018,34(11):3217-3224.
LIU Yanpeng, ZHU Lixin, ZHOU Yongzhang. Application of Convolutional Neural Network in prospecting prediction of ore deposits: Taking the Zhaojikou Pb-Zn ore deposit in Anhui Province as a case[J]. Acta Petrologica Sinica, 2018, 34 (11): 3217-3224.
[90] 多南訊,呂強,林輝燦,等.邁進(jìn)高維連續(xù)空間:深度強化學(xué)習(xí)在機器人領(lǐng)域中的應(yīng)用[J].機器人,2019,41(2):142-154.
DUO Nanxun, Lü Qiang, LIN Huican, et al. Step into high-dimensional and continuous action space: A survey on applications of deep reinforcement learning to robotics[J]. Robot, 2019, 41 (2): 142-154.
[91] Mnih V, Kavukcuoglu K, Silver D, et al. Playing Atari with deep reinforcement learning[EB/OL]. NIPS Deep Learning Workshop, 2013. https://arxiv.org/abs/1312.5602.
[92] Deisenroth M P, Rasmussen C E. PILCO: A model-based and data-efficient approach to policy search[C]// 28thInternational Conference on Machine Learning. New York, USA: ACM: 2011: 465-472.
[93] 徐述騰,周永章.基于深度學(xué)習(xí)的鏡下礦石礦物的智能識別實驗研究[J].巖石學(xué)報,2018,34(11):3244-3252.
XU Shuteng, ZHOU Yongzhang. Artificial intelligence identification of ore minerals under microscope based on deep learning algorithm[J]. Acta Petrologica Sinica, 2018, 34 (11): 3244-3252.
[94] 石磊.機器人與三維激光掃描儀在礦井地質(zhì)測量中的聯(lián)合應(yīng)用[J].機械管理開發(fā),2020,7:199-200.
SHI Lei. Combined application of robot and 3D laser scanner in mine geological survey[J]. Mechanical Management and Development, 2002, 7:199-200.
[95] 侯遠(yuǎn)韶.基于改進(jìn)蟻群算法在機器人路徑優(yōu)化中的應(yīng)用[J].安陽工學(xué)院學(xué)報,19(6):39-42.
HOU Yuanshao. 2002. Application of improved ant colony algorithm in robot path optimization[J]. Journal of Anyang Institute of Technology, 2020, 19 (6): 39-42.
[96] 馬靜雅,張樹生,姚春清,等.礦用軌道機器人測速度、加速度、位移裝置及方法[J].煤炭技術(shù),2019,38(1):167-168.
MA Jingya, ZHANG Shusheng, YAO Chunqing, et al. Mine orbiting robot measuring speed, acceleration and displacement device and method[J]. Coal Technology, 2019, 38 (11): 167-168.
[97] 姜妍,張立國.深度學(xué)習(xí)的對抗攻擊與防御方法綜述[J].計算機工程,2020,https://doi.org/10.19678/j.issn.1000-3428.0059156.
JIANG Yan, ZHANG Liguo. Survey of adversarial attacks and defense methods of deep learning[J]. Computer Engineering, 2020, https://doi.org/10.19678/j.issn.1000-3428.0059156.
[98] Tesauro G.TD-gammon, a self-teaching backgammon program,achieves master-level play[J]. Neural Computation, 1994, 6 (2): 215-219.
[99] Kocsis L, Szepesvári C. Bandit based Monte-Carlo planning[M]. Lecture Notes in Computer Science, Vol.4212. Berlin: Springer, 2006: 282-293.
[100]Namatēvs I. Deep convolutional neural networks: Structure, feature extraction and training[J]. Information Technology and Management Science, 2017, 20 (1): 40-47.
[101] 施恩,李騫,顧大權(quán).基于局部特征的卷積神經(jīng)網(wǎng)絡(luò)模型[J].計算機工程,2018,44(2):282-286.
SHI En, LI Qian, GU Daquan. Convolutional neural network nodel nased on nocal feature[J]. Computer Engineering, 2018, 44(2): 282-286.
[102] Bianco S, Buzzelli M, Mazzini D, et al. Deep learning for logo recognition[J]. Neurocomputing, 2017, 245: 23-30.
[103] Riedmiller M. Neural fitted Q iteration – First experiences with a data efficient neural reinforcement learning method[C]// 16thEuropean Conference on Machine Learning. Berlin, Germany: Springer-Verlag, 2005: 317-328.
[104] Farahmand A M, Nabi S, Nikovski D N. Deep reinforcement learning for partial differential equation control[C]// American Control Conference. Piscataway, USA: IEEE, 2017: 3120-3127.
[105] 周永章,陳川,張旗,等.地質(zhì)大數(shù)據(jù)分析的若干工具與應(yīng)用[J].大地構(gòu)造與成礦學(xué),2020,44(2):173-182.
ZHOU Yongzhang, CHEN Chuan, ZHANG Qi, et al. Introduction of tools for geological big data mining and their applications[J]. Geotectonica et Metallogenia, 2020, 44 (2): 173-182.
[106] Lum P Y, Singh G, Lehman A, et al. Extracting insights from the shape of complex data using topology[J]. Scientific Reports, 2013, 3: 1236.
[107] 周永章,陳爍,張旗,等.大數(shù)據(jù)與數(shù)學(xué)地球科學(xué)研究進(jìn)展:大數(shù)據(jù)與數(shù)學(xué)地球科學(xué)專題代序[J].巖石學(xué)報,2018,34(2):255 -263.
ZHOU Yongzhang, CHEN Shuo, ZHANG Qi, et al. Advances and prospects of big data and mathematical geoscience[J]. Acta Petrologica Sinica, 2018, 34 (2): 255-263.
[108] Liu H, Sun W D, Zartman N R, et al. Continuous plate subduction marked by the rise of alkali magmatism 2.1 billion years ago[J]. Nature Communications, 2019, 10: 3408.
[109] 常力恒,朱月琴,張戈一,等.面向礦產(chǎn)資源信息的空間關(guān)聯(lián)性分析[J].巖石學(xué)報,2018,34(2):314-318.
CHANG Liheng, ZHU Yueqin, ZHANG Geyi, et al. Spatial correlation analysis of mineral resources information[J]. Acta Petrologica Sinica, 2018, 34 (2): 314-318.
[110] 周永章,黎培興,王樹功,等.礦床大數(shù)據(jù)及智能礦床模型研究背景與進(jìn)展[J].礦物巖石地球化學(xué)通報,2017,36(2):327-331,344.
ZHOU Yongzhang, LI Peixing, WANG Shugong, et al. Researchprogress on big data and intelligent modelling of mineral deposits[J]. Bulletin of Mineralogy, Petrology and Geochemistry, 2017, 36 (2): 327-331, 344.
[111] Richards J P, Spell T, Rameh E, et al. High Sr/Y magmas reflect are magmatic water content, and porphyry Cu±Mo±Au potential: examples from the Tethyan arcs of central and eastern Iran and western Pakistan[J]. Economic Geology, 2012,107 (2):295-332.
[112] 陳進(jìn),毛先成,劉占坤,等.基于隨機森林算法的大尹格莊金礦床三維成礦預(yù)測[J].大地構(gòu)造與成礦學(xué),2020,44(2):231-241.
CHEN Jin, MAO Xiancheng, LIU Zhankun, et al. Three-dimensional metallogenic prediction based on random forest classification algorithm for the Dayingezhuang gold deposit[J]. Geotectonica et Metallogenia, 2020, 44 (2): 231-241.
[113] 劉艷鵬,朱立新,周永章.大數(shù)據(jù)挖掘與智能預(yù)測找礦靶區(qū)實驗研究:卷積神經(jīng)網(wǎng)絡(luò)模型的應(yīng)用[J].大地構(gòu)造與成礦學(xué),2020,44(2):192-202.
LIU Yanpeng, ZHU Lixin, ZHOU Yongzhang. Experimental research on big data mining and intelligent prediction of prospecting target area: Application of convolutional neural network model[J]. Geotectonica et Metallogenia, 2020, 44(2): 192-202.
[114] 閆巖,陳軍林,彭潤民,等.變質(zhì)作用對沉積巖容礦鉛鋅礦床的改造與成礦效應(yīng):數(shù)據(jù)挖掘的啟示[J].礦物學(xué)報, 2021,https://doi.org/10.16461/j.cnki.1000-4734.2020.41.059.
YAN Yan, CHEN Junlin, PENG Runmin, et al. Metamorphosed overprints on the sediment-hosted zinc-lead deposits:revelation from data mining[J]. Acta Mineralogica Sinica, 2021, https://doi.org/10.16461/j.cnki.1000-4734.2020.41.059.
[115] 張振杰,成秋明,楊玠,等.機器學(xué)習(xí)與成礦預(yù)測:以閩西南鐵多金屬礦預(yù)測為例[J/OL].地學(xué)前緣,2021,https://doi.org/10.13745/j.esf.sf.2021.1.4.
ZHANG Zhenjie, CHENG Qiuming, YANG Jie, et al. Machine learning for prospectivity: a case study of iron-polymetallic mineral prospectivity in Southwestern Fujian[J/OL]. Earth Science Frontiers, 2021, https://doi.org/10.13745/j.esf.sf.2021.1.4.
[116] 張夏林,吳沖龍,周琦,等.基于勘查大數(shù)據(jù)和數(shù)據(jù)集市的錳礦床三維地質(zhì)建模[J].地質(zhì)科技通報,2020,39(4):12-20.
ZHANG Xialin, WU Chonglong, ZHOU Qi, et al. Three-dimensional geological modeling of manganese deposits based on exploration big data and data market[J]. Bulletin of Geological Science and Technology, 2020, 39(4): 12-20.
[117] 奚小環(huán).大數(shù)據(jù)科學(xué)從信息化、模式化到智能化:現(xiàn)代地球化學(xué)應(yīng)用研究的新范式[J].地學(xué)前緣(中國地質(zhì)大學(xué)(北京);北京大學(xué)),2021,28(1):308-317.
XI Xiaohuan. Big data science from informationization to modelling to intelligentization: new paradigm of applied geochemical research[J]. Earth Science Frontiers, 2021, 28(1): 308-317.
[118] 洪雙,左仁廣,胡浩,等.磁鐵礦元素地球化學(xué)大數(shù)據(jù)構(gòu)建及其在礦床成因分類中的應(yīng)用[J/OL].地學(xué)前緣,2021,https://doi.org/10.13745/j.esf.sf.1.10.
HONG Shuang, ZUO Renguang, HU Hao, et al. Construction of magnetite geochemical big data and its application in genetic classification of ore deposits[J/OL]. Earth Science Frontiers, 2021, https://doi.org/10.13745/j.esf.sf.1.10.
[119] 馬瑤,趙江南.機器學(xué)習(xí)方法在礦產(chǎn)資源定量預(yù)測應(yīng)用研究進(jìn)展[J].地質(zhì)科技通報,2021,40(1):132-141.
MA Yao, ZHAO Jiangnan. Advances in the application of machine learning methods in mineral prospectivity mapping[J]. Bulletin of Geological Science and Technology, 2021, 40 (1): 132-141.
[120] 吳沖龍,劉剛,周琦,等.地質(zhì)科學(xué)大數(shù)據(jù)統(tǒng)合應(yīng)用的基本問題[J].地質(zhì)科學(xué)通報,2020,39(4):1-11.
WU Chonglong, LIU Gang, ZHOU Qi, et al. Fundamental problems of integrated application of big data in geoscience[J]. Bulletin of Geological Science and Technology,2020, 39 (4): 1-11.
[121] Wikipedia Foundation. Inc. Data lake[EB/OL]. https://en. wikpedia. Org/wiki/Data_lake[2018-07-16](2019-11-18).
[122] 崔文泉,黃禹僑.高維數(shù)據(jù)情形下的一種基于隨機投影的集成分類方法[J].中國科學(xué)技術(shù)大學(xué)學(xué)報,2019,49(12):974-984.
CUI Wenquan, HUANG Yuqiao. A new random projection-based ensemble classifier for high-dimensional data[J]. Journal of University of Science and Technology of China, 2019, 49 (12):974-984.
[123] Roweis S T, Saul L K. Nonlinear dimensionality reduction by locally linear embedding[J]. Science, 2020, 290 (5500): 2323-2326.
[124] 顏丙云,于飛,黃彪.基于即時學(xué)習(xí)的不完整高維數(shù)據(jù)非線性過程軟測量[J].青島科技大學(xué)學(xué)報(自然科學(xué)版),2020,41(5):109-118.
YAN Bingyun,YU Fei,HUANG Biao. Nonlinear process soft sensors of incomplete high-dimensional data based on just-in-time learning[J]. Joural of Qingdao University of Science and Technology (Natural Science Edition), 2020, 41 (5): 109-118.
[125] 杜杰,王驍,胡良劍.非線性降維技術(shù)與可視化應(yīng)用[J].東華大學(xué)學(xué)報(自然科學(xué)版),2020,46(4):675-680.
DU Jie, WANG Xiao, HU Liangjian. Technology of nonlinear dimension reduction and its application in visualization[J].Journal of Donghua University(Natural Science Edition), 2020, 46 (4): 675-680.
[126] 薛紅新.基于機器學(xué)習(xí)方法的分類與預(yù)測問題研究[D].太原:中北大學(xué),2019,1-147.
XUE Hongxin. Research on the problems of classification and prediction based on machine learning methods[D]. Taiyuan: North University of China, 2019, 1-147.
[127] Babcock B, Babu S, Datar M, et al. Models and is sues in data streams[C]∥Proceedings of the 21stACM SIGACT-SIGMOD SIGART Symposium on Principles of Database Systems. Madison, Wisconsin: ACM Press, 2002, 1-16.
[128] 李國徽,陳輝,楊兵,等.基于概率模型的數(shù)據(jù)流預(yù)測查詢算法[J].計算機科學(xué),2008,35(4):66-69.
LI Guohui, CHEN Hui, YANG Bing, et al. Predictive queries algorithm based on probability model over data streams[J]. Compoter Science, 2008, 35 (4): 66-69.
A review on the development of mineral deposit science in the era of big data
LU Yingchuan1,LI Peng1,WANG Hao1,ZHANG Xiang1,TANG Yulei1,XIE Gen2
(1. Center of Geophysics Survey,CGS,Langfang,Hebei 065000,China;2. Langfang Center for Integrated Natural Resources Survey,China Geological Survey,Langfang,Hebei 065000,China)
In recent years, there have been some bottlenecks in the study of mineral deposits,such as. the innovation deficienciences of metallogenic model and the monotorous perspective of metallogenic series and metallogeny. This paper reviewed the development history of ore deposit science and pointed out that every breakthrough and leap in ore deposit science is closely related to the development of new science and technology. With the continuous improvement of scientific progress, especially the coming of the ‘big data’ and ‘intelligent’ era, new technologies such as deep learning of big data in artificial intelligence are developing vigorously. Geological big data has the characteristics of ‘4V’ of ‘big data’, such as ‘volume’,‘velocity’,‘variety’ and ‘value’,as well as the characteristics of pluralism, multi-dimension, multi-source, heterogeneity and space- time. Based on the statistical comparison and analysis of relevant domestic and foreign literatures in the past decade, this paper expounded the subordinate relationship and main characteristics among artificial intelligence, machine learning and deep learning, and the examples of random forest algorithm, convolutional neural network, decision tree algorithm, Naive Bayes algorithm and support vector machine and other algorithms in ore deposit research were also sorted out. In this paper, it was considered that the study of intelligent exploration of global mineral resources that lead by the artificial intelligence technology will become the inevitable direction of the future development of ore deposit science.
mineralogy;metallogenic model;big data;machine learning;artificial intelligence
P61;TP31
A
1672-0636 (2021) 03-0295-16
10.3969/j.issn.1672-0636.2021.03.002
中國地質(zhì)調(diào)查局項目(編號:DD20191023)資助。
2021-04-13
路英川(1986— ),男,河北邯鄲人,博士,從事礦物學(xué)、巖石學(xué)、礦床學(xué)研究工作。E-mail:luyingchuan2008@163.com
Supported by the China Geological Survey project (No. DD20191023).
LU Yingchuan,born in 1986,doctor, mainly engages in research on mineral deposits.