機(jī)器學(xué)習(xí)與物理專題編者按

機(jī)器學(xué)習(xí), 尤其是深度學(xué)習(xí), 在很多方面取得了令人矚目的成就, 是當(dāng)前科學(xué)技術(shù)領(lǐng)域最為熱門、發(fā)展最快的方向之一. 其與物理的結(jié)合是最近幾年新興的交叉前沿領(lǐng)域, 受到了廣泛關(guān)注. 一方面, 運(yùn)用機(jī)器學(xué)習(xí)的方法可以解決一些復(fù)雜的、傳統(tǒng)方法很難或無法解決的物理問題; 另一方面, 物理中的一些概念、理論和方法也可以用于研究機(jī)器學(xué)習(xí). 二者的交叉融通帶來了新的機(jī)遇與挑戰(zhàn),將極大地促進(jìn)兩個(gè)領(lǐng)域的發(fā)展.

本專題邀請(qǐng)了若干活躍在該新興領(lǐng)域的專家撰稿, 重點(diǎn)介紹機(jī)器學(xué)習(xí)與物理交叉方向的部分國(guó)際前沿課題和最新研究進(jìn)展. 內(nèi)容涵蓋了量子人工智能中的對(duì)抗學(xué)習(xí), 量子生成模型, 基于波動(dòng)與擴(kuò)散的機(jī)器學(xué)習(xí), 自動(dòng)微分, 絕熱量子算法設(shè)計(jì), 量子機(jī)器學(xué)習(xí)中的編碼與初態(tài)制備, 以及基于自旋體系的量子機(jī)器學(xué)習(xí)實(shí)驗(yàn)進(jìn)展等.

希望本專題能夠幫助讀者了解機(jī)器學(xué)習(xí)與物理交叉方向的研究?jī)?nèi)容, 基本思想與方法, 最新進(jìn)展情況, 以及面臨的挑戰(zhàn)與機(jī)遇. 同時(shí), 也希望這個(gè)專題能夠激發(fā)讀者的興趣, 吸引更多的研究人員加入到此交叉領(lǐng)域的研究中.

(客座編輯: 鄧東靈 清華大學(xué))

Machine learning, especially deep learning, has achieved remarkable success in a wide range of applications. It is one of today’s most rapidly growing fields in science and technology. In recent years, the interplay between machine learning and physics has attracted tremendous attention, giving rise to a new interdisciplinary research frontier. On the one hand, we may utilize machine learning methods to tackle certain intricate physical problems that are beyond the capability of traditional approaches. On the other hand, certain concepts, ideas, and methods originated in physics can also be exploited to enhance the study of machine learning.Without a doubt, the fusion of machine learning and physics will bring us new opportunities and challenges, and significantly advance the studies in both fields.

This special topic contains several review papers written by experts working actively in this emergent interdisciplinary field. These papers review a number of hot topics and some latest progresses, covering adversarial learning in quantum artificial intelligence, quantum generative models, machine learning based on waves and diffusions, automatic differentiation, machine learning assisted quantum adiabatic algorithm design, state preparation in quantum machine learning, experimental progress of quantum machine learning based on spin systems, etc.

We hope this special topic can help readers gain a primary picture of the research content,basic ideas and methods, the latest developments, and the challenges and opportunities faced in the intersection of machine learning and physics. Meanwhile, we also hope this special topic can provide some inspiration to readers, and attract more researchers to join this exciting interdisciplinary field.

Deng Dong-Ling