Review of Application of Deep Learning in Symbolic Music Generation

doi:10.3778/j.issn.1002-8331.2209-0305

Abstract

Abstract: Symbolic music generation is an important task in the field of music information retrieval. This paper provides a comprehensive summary of deep learning-based symbolic music generation, and existing methods are classified, analyzed, as well as compared. The research status and tasks of symbolic music generation are introduced in detail. It expounds the representation and coding methods of symbolic music, and focuses on the induction, comparison and analysis of deep learning-based models, which are divided into three categories according to different basic structures. It expounds and summarizes the evaluation criteria and datasets in the field of symbolic music generation, and evaluates the performance of representative models. The existing problems in this field are pointed out and corresponding prospects are put forward.

Key words: artificial intelligence, symbolic music, intelligent composing, deep learning, neural network

摘要： 符号音乐生成是音乐信息检索领域中的一个重要任务。对基于深度学习的符号音乐生成进行了全面总结，并对已有方法进行分类、分析和比较。详细介绍了符号音乐生成研究现状及其任务。阐述符号音乐表征及编码方法，并重点对基于深度学习的模型进行归纳比较与分析，根据不同的基础架构分为三类。阐述并归纳符号音乐生成领域的评价标准及数据集等资源，对代表性模型的性能进行评估对比。指出该领域目前存在的问题并提出相应的展望。

关键词: 人工智能, 符号音乐, 智能作曲, 深度学习, 神经网络

CHEN Jishang, Abudukelimu Halidanmu, LIANG Yunze, Abulizi Abudukelimu, Aishan Mikelayi, GUO Wenqiang. Review of Application of Deep Learning in Symbolic Music Generation[J]. Computer Engineering and Applications, 2023, 59(9): 27-45.

陈吉尚, 哈里旦木·阿布都克里木, 梁蕴泽, 阿布都克力木·阿布力孜, 米克拉依·艾山, 郭文强. 深度学习在符号音乐生成中的应用研究综述[J]. 计算机工程与应用, 2023, 59(9): 27-45.

References

[1] FANG Y，XU Y，LI H，et al.Writing in the air：recognize letters using deep learning through WiFi signals[C]//2020 6th International Conference on Big Data Computing and Communications（BIGCOM），2020：8-14.
[2] CHEN S.Exploration of artistic creation of Chinese ink style painting based on deep learning framework and convolutional neural network model[J].Soft Computing，2020，24（11）：7873-7884.
[3] YANG Y H.Automatic music composition with transformers[C]//Proceedings of the 2021 International Joint Workshop on Multimedia Artworks Analysis and Attractiveness Computing in Multimedia，2021.
[4] PRASHANT-KRISHNAN V，RAJARAJESWARI S，KRISHN-AMOHAN V，et al.Music generation using deep learning techniques[J].Journal of Computational and Theoretical Nanoscience，2020，17（9/10）：3983-3987.
[5] DONAHUE C，MAO H H，LI Y E，et al.LakhNE-S：improving multi-instrumental music generation with cross-domain pre-training[C]//International Society for Music Information Retrieval，2019：685-692.
[6] SHENG Z，SONG K，TAN X，et al.Songmass：automatic song writing with pre-training and alignment constraint[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2021：13798-13805.
[7] BRUNNER G，KONRAD A，WANG Y，et al.MIDI-VAE：modeling dynamics and instrumentation of music with applications to style transfer[C]//International Society for Music Information Retrieval，2018：747-754.
[8] PATI A，LERCH A，HADJERES G.Learning to traverse latent spaces for musical score inpainting[C]//International Society for Music Information Retrieval，2019：343-351.
[9] BRIOT J P，PACHET F.Deep learning for music generation：challenges and directions[J].Neural Computing and Applications，2020，32（4）：981-993.
[10] HERNANDEZ-OLIVAN C，BELTRAN J R.Music composition with deep learning：a review[J].arXiv：2108. 12290，2021.
[11] PINKERTON R C.Information theory and melody[J].Scientific American，1956，194（2）：77-87.
[12] HILLER L A，ISAACSON L M.Experimental music；composition with an electronic computer[M].[S.l.]：Greenwood Publishing Group Inc，1979：197.
[13] PUTNAM J.A grammar-based genetic programming technique applied to music generation[C]//Evolutionary Programming，1996：363-368.
[14] VAN DER MERWE A，SCHULZE W.Music generation with Markov models[J].IEEE Multimedia，2010，18（3）：78-85.
[15] SHAO X，XU C，KANKANHALLI M S.Unsupervised classification of music genre using hidden Markov model[C]//2004 IEEE International Conference on Multimedia and Expo（ICME），2004：2023-2026.
[16] MOZER M C.Neural network music composition by prediction：exploring the benefits of psychoacoustic constraints and multi-scale processing[J].Connection Science，1994，6（2/3）：247-280.
[17] HORI T，WADA S，TAI H，et al.Automatic music score recognition/play system based on decision based neural network[C]//1999 IEEE Third Workshop on Multimedia Signal Processing，1999：183-184.
[18] SIGTIA S，DIXON S.Improved music feature learning with deep neural networks[C]//2014 IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2014：6959-6963.
[19] TODD P M.A connectionist approach to algorithmic composition[J].Computer Music Journal，1989，13（4）：27-43.
[20] ELMAN J L.Finding structure in time[J].Cognitive Science，1990，14（2）：179-211.
[21] LATTNER S，GRACHTEN M，AGRES K，et al.Probabilistic segmentation of musical sequences using restricted Boltzmann machines[C]//International Conference on Mathematics and Computation in Music.Cham：Springer，2015：323-334.
[22] CHEN K，ZHANG W，DUBNOV S，et al.The effect of explicit structure encoding of deep neural networks for symbolic music generation[C]//2019 International Workshop on Multilayer Music Representation and Processing（MMRP），2019：77-84.
[23] LO M Y，LUCAS S M.Evolving musical sequences with n-gram based trainable fitness functions[C]//2006 IEEE International Conference on Evolutionary Computation，2006：601-608.
[24] GOEL K，VOHRA R，SAHOO J K.Polyphonic music generation by modeling temporal dependencies using a RNNDBN[C]//International Conference on Artificial Neural Networks.Cham：Springer，2014：217-224.
[25] HOCHREITER S，SCHMIDHUBER J.Long short-term memory[J].Neural Computation，1997，9（8）：1735-1780.
[26] BRETAN M，WEINBERG G，HECK L P.A unit selection methodology for music generation using deep neural networks[C]//Proceedings of the Eighth International Conference on Computational Creativity，2017：19-23.
[27] KINGMA D P，WELLING M.Auto-encoding variational bayes[J].arXiv：1312.6114v9，2014.
[28] VASWANI A，SHAZEER N，PARMAR N，et al.Attention is all you need[C]//Advances in Neural Information Processing Systems，2017：30.
[29] GOODFELLOW I，POUGET-ABADIE J，MIRZA M，et al.Generative adversarial nets[C]//Advances in Neural Information Processing Systems，2014：27.
[30] ROBERTS A，ENGEL J，RAFFEL C，et al.A hierarchical latent vector model for learning long-term structure in music[C]//International Conference on Machine Learning，2018：4364-4373.
[31] JIANG J，XIA G G，CARLTON D B，et al.Transformer VAE：a hierarchical model for structureaware and interpretable music representation learning[C]//2020 IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2020：516-520.
[32] WANG Z，ZHANG Y，ZHANG Y，et al.Pianotree VAE：structured representation learning for poly-phonic music[J].arXiv：2008.07118，2020.
[33] ECK D，SCHMIDHUBER J.A first look at music composition using LSTM recurrent neural networks[J].Istituto Dalle Molle Di Studi Sull Intelligenza Artificiale，2002，103：48.
[34] LIM H，RHYU S，LEE K.Chord generation from symbolic melody using blstm networks[C]//International Society for Music Information Retrieval，2017：621-627.
[35] YANG W，SUN P，ZHANG Y，et al.Clstms：a combination of two LSTM models to generate chords accompaniment for symbolic melody[C]//2019 International Conference on High Performance Big Data and Intelligent Systems（HPBD&IS），2019：176-180.
[36] CHEN Z，YU J M.GCA：a chord music generation algorithm based on double-layer LSTM[C]//2021 3rd International Conference on Advances in Computer Technology，Information Science and Communication（CTISC），2021：57-61.
[37] WU S，YANG Y，WANG Z，et al.Melody harmonization with controllable harmonic rhythm[J].arXiv：2112.11122，2021.
[38] GATYS L A，ECKER A S，BETHGE M.Image style transfer using convolutional neural networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：2414-2423.
[39] YU K，XU W，GONG Y.Transfer learning methods and systems for feed-forward visual recognition systems：U.S.Patent 8，345，962[P].2013-01-01.
[40] HUNG H T，WANG C Y，YANG Y H，et al.Improving automatic jazz melody generation by transfer learning techniques[C]//2019 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference（APSIPA ASC），2019：339-346.
[41] MEREDITH D.Method of computing the pitch names of notes in MIDI-like music representations：U.S.Patent Application 10/821，962[P].2004-11-04.
[42] BOULANGER-LEWANDOWSKI N，BENGIO Y，VINCENT P.Modeling temporal dependencies in high dimensional sequences：application to polyphonic music generation and transcription[C]//Proceedings of the 29th International Coference on International Conference on Machine Learning，June 2012：1881-1888.
[43] BAGGI D，HAUS G.IEEE 1599：music encoding and interaction[J].Computer，2009，42（3）：84-87.
[44] KEPPER J.A data model for digital musicology and its current state-the music encoding initiative[C]//Digital Humanities，2010：184
[45] THAKKER U，DASIKA G，BEU J，et al.Measuring scheduling efficiency of rnns for NLP applications[J].arXiv：1904.03302，2019.
[46] SIGTIA S，BENETOS E，CHERLA S，et al.An RNN-based music language model for improving automatic music transcription[C]//International Society for Music Information Retrieval，2014：53-58.
[47] SUBAKAN Y C，SMARAGDIS P.Diagonal RNNs in symbolic music modeling[C]//2017 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics（WASPAA），2017：354-358.
[48] STURM B，SANTOS J F，BEN-TAL O，et al.Music transcription modelling and composition using deep learning[C]//1st Conference on Computer Simulation of Musical Creativity，2016.
[49] LIANG F T，GOTHAM M，JOHNSON M，et al.Automatic stylistic composition of bach chorales with deep LSTM[C]//International Society for Music Information Retrieval，2017：449-456.
[50] MAO H H，SHIN T，COTTRELL G.DeepJ：style specific music generation[C]//2018 IEEE 12th International Conference on Semantic Computing（ICSC），2018：377-382.
[51] CHEN K，XIA G，DUBNOV S.Continuous melody generation via disentangled short-term representations and structural conditions[C]//2020 IEEE 14th International Conference on Semantic Computing（ICSC），2020：128-135.
[52] ZHAO T，ZHAO R，ESKENAZI M.Learning discourse-level diversity for neural dialog models using conditional variational autoencoders[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics（Volume 1：Long Papers），2017：654-664.
[53] AHMED K，KESKAR N S，SOCHER R.Weighted transformer network for machine translation[J].arXiv：1711. 02132，2017.
[54] LI Y，FENG R，REHG I，et al.Transformer-based neural text generation with syntactic guidance[J].arXiv：2010. 01737，2020.
[55] HUANG C Z A，VASWANI A，USZKOREIT J，et al.Music transformer[C]//International Conference on Learning Representations，2018.
[56] SHAW P，USZKOREIT J，VASWANI A.Self-attention with relative position representations[C]//Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics：Human Language Technologies（Volume 2：Short Papers），2018：464-468.
[57] DAI Z，YANG Z，YANG Y，et al.Transformer-xl：attentive language models beyond a fixed-length context[C]//Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics，2019：2978-2988.
[58] CHILD R，GRAY S，RADFORD A，et al.Generating long sequences with sparse transformers[J].arXiv：1904. 10509，2019.
[59] CHOI K，HAWTHORNE C，SIMON I，et al.Encoding musical style with transformer autoencoders[C]//International Conference on Machine Learning，2020：1899-1908.
[60] VON RüTTE D，BIGGIO L，KILCHER Y，et al.FIG-ARO：generating symbolic music with finegrained artistic control[J].arXiv：2201.10936，2022.
[61] HUANG Y S，YANG Y H.Pop music transformer：beat-based modeling and generation of expressive pop piano compositions[C]//Proceedings of the 28th ACM International Conference on Multimedia，2020：1180-1188.
[62] HSIAO W Y，LIU J Y，YEH Y C，et al.Compound word transformer：learning to compose full song music over dynamic directed hypergraphs[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2021：178-186.
[63] KATHAROPOULOS A，VYAS A，PAPPAS N，et al.Transformers are RNNS：fast autoregressive transformers with linear attention[C]//International Conference on Machine Learning，2020：5156-5165.
[64] ENS J，PASQUIER P.Mmm：exploring conditional multi-track music generation with the transformer[J].arXiv：2008.06048，2020.
[65] DONG H W，CHEN K，DUBNOV S，et al.Multitrack music transformer：learning long-term dependencies in music with diverse instruments[J].arXiv：2207.06983，2022.
[66] WU J，LIU X，HU X，et al.PopMNet：generating structured pop music melodies using neural networks[J].Artificial Intelligence，2020，286：103303.
[67] ZOU Y，ZOU P，ZHAO Y，et al.MELONS：generating melody with long-term structure using transformers and structure graph[C]//2022 IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2022：191-195.
[68] NASH J.Non-cooperative games[J].Annals of Mathematics，1951：286-295.
[69] HUANG H，YU P S，WANG C.An introduction to image synthesis with generative adversarial nets[J].arXiv：1803. 04469，2018.
[70] YU L，ZHANG W，WANG J，et al.Seqgan：sequence generative adversarial nets with policy gradient[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2017.
[71] MOGREN O.C-RNN-GAN：continuous recurent neural networks with adversarial training[C]//Conference and Workshop on Neural Information Processing Systems，2018.
[72] TRIEU N，KELLER R M.JazzGAN：improvising with generative adversarial networks[C]//6th International Workshop on Musical Metacreation，2018.
[73] OORD A，DIELEMAN S，ZEN H，et al.Wavenet：a generative model for raw audio[C]//The 9th ISCA Speech Synthesis Workshop，Sunnyvale，CA，USA，13-15 September，2016：125.
[74] YANG L C，CHOU S Y，YANG Y H.MidiNet：a convolutional generative adversarial network for symbolic-domain music generation[C]//International Society for Music Information Retrieval，2017：324-331.
[75] WAITE E.Generating long-term structure in songs and stories[J].Web Blog Post，2016，15（4）.
[76] DONG H W，HSIAO W Y，YANG L C，et al.Musegan：multi-track sequential generative adversarial networks for symbolic music generation and accompaniment[C]//Proceedings of the AAAI Conference on Artificial Intelligence，2018.
[77] 汪涛，靳聪，李小兵，等.基于Transformer的多轨音乐生成对抗网络[J].计算机应用，2021，41（12）：3585-3589.
WANG T，JIN C，LI X B，et al.Multi-track music generative adversarial network based on Transformer[J].Journal of Computer Applications，2021，41（12）：3585-3589.
[78] 白勇，帖云，靳聪，等.基于强化学习的智能作曲研究[J].人工智能，2020（2）：47-56.
BAI Y，TIE Y，JIN C，et al.Research on intelligent composition based on reinforcement learning[J].Artificial Intelligence，2020（2）：47-56.
[79] GUL O，SCHLAGER C，TODD G.MuML：musical meta-learning[EB/OL].（2021-10-01）[2023-04-11].https：//vdocuments.mx/muml-musical-meta-learning-stanford-university.html?page=1.
[80] KOTECHA N.Bach2bach：generating music using a deep reinforcement learning approach[J].arXiv：1812.01060，2018.
[81] XUE L，SONG K，WU D，et al.DeepRapper：neural rap generation with rhyme and rhythm modeling[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing（Volume 1：Long Papers），2021：69-81.
[82] LIU J，DONG Y，CHENG Z，et al.Symphony generation with permutation invariant language model[J].arXiv：2205.05448，2022.
[83] SIMON I，OORE S.Performance RNN：generating music with expressive timing and dynamics[EB/OL].（2017-07-29）.https：//magenta.tensorflow.org/performance-rnn.
[84] HADJERES G，PACHET F，NIELSEN F.Deepbach：a steerable model for bach chorales generation[C]//International Conference on Machine Learning，2017：1362-1371.
[85] DONG H，YANG Y.Convolutional generative adversarial networks with binary neurons for polyphonic music generation[C]//Proceedings of the 19th International Society for Music Information Retrieval Conference， 2018：190-196.
[86] PAYNE C.Musenet[EB/OL].（2019-08-25）.https：//openai.com/research/musenet.
[87] HADJERES G，NIELSEN F.Anticipation-RNN：enforcing unary constraints in sequence generation，with application to interactive music generation[J/OL].Neural Computing & Applications，2020：995?1005[2022?09?08].https：//doi.org/10.1007/s00521-018-3868-4.
[88] REN Y，HE J，TAN X，et al.Popmag：pop music accompaniment generation[C/OL]//Proceedings of the 28th ACM International Conference on Multimedia，2020：1198-1206[2022-09-08].https：//doi.org/10.1145/3394171.3413721.
[89] WU S L，YANG Y H.MuseMorphose：full-song and fine-grained piano music style transfer with one Transformer VAE[J].arXiv：2105.04090，2021.
[90] SHIH Y J，WU S L，ZALKOW F，et al.Theme transformer：symbolic music generation with theme-conditioned transformer[J].IEEE Transactions on Multimedia，2022.
[91] GU Y，YIN X，RAO Y，et al.Bytesing：a Chinese singing voice synthesis system using duration allocated encoder-decoder acoustic models and wavernn vocoders[C]//2021 12th International Symposium on Chinese Spoken Language Processing（ISCSLP），2021：1-5.
[92] YANG L C，LERCH A.On the evaluation of generative models in music[J].Neural Computing and Applications，2020，32（9）：4773-4784.
[93] BONNIN G，JANNACH D.Automated generation of music playlists：survey and experiments[J].ACM Computing Surveys（CSUR），2014，47（2）：1-35.
[94] CAMPBELL M.Objective evaluation of musical instrument quality：a grand challenge in musical acoustics[J].The Journal of the Acoustical Society of America，2013，19（1）：032003.
[95] GODWIN T，RIZOS G，BAIRD A，et al.Evaluating deep music generation methods using data augmentation[C]//2021 IEEE 23rd International Workshop on Multimedia Signal Processing（MMSP），2021：1-6.
[96] GARCIA-VALENCIA S，BETANCOURT A，LALINDE-PULIDO J G.A framework to compare music generative models using automatic evaluation metrics extended to rhythm[J].arXiv：2101.07669，2021.
[97] RIBEIRO F，FLORêNCIO D，ZHANG C，et al.Crowdmos：an approach for crowdsourcing mean opinion score studies[C]//2011 IEEE International Conference on Acoustics，Speech and Signal Processing（ICASSP），2011：2416-2419.
[98] TURING A M，HAUGELAND J.Computing machinery and intelligence[J].The Turing Test：Verbal Behavior as the Hallmark of Intelligence，1950：29-56.
[99] SCHOFIELD J.Computer chatbot “Eugene Goostman” passes the turing test[EB/OL].[2014-06].https：//www.zdnet.com/article/computer-chatbot-eugene-goostmanpasses-the-turing-test.
[100] KONG M，HUANG L.Bach style music authoring system based on deep learning[J].arXiv：2110.02640，2021.
[101] CíFKA O，?IM?EKLI U，RICHARD G.Supervised symbolic music style translation using synthetic data[C]//International Society for Music Information Retrieval，2019：588-595.
[102] LIU I，RAMAKRISHNAN B.Bach in 2014：music composition with recurrent neural network[J].arXiv：1412.3191，2014.
[103] ARIZA C.The interrogator as critic：the turing test and the evaluation of generative music systems[J].Computer Music Journal，2009，33（2）：48-70.
[104] BECH S，ZACHAROV N.Perceptual audio evaluation-theory，method and application[M].[S.l.]：John Wiley & Sons，2007.
[105] HERNANDEZ-OLIVAN C，PUYUELO J A，BELTRAN J R.Subjective evaluation of deep learning models for symbolic music composition[J].arXiv：2203.14641，2022.
[106] JELINEK F，MERCER R L，BAHL L R，et al.Perplexity—a measure of the difficulty of speech recognition tasks[J].The Journal of the Acoustical Society of America，1977，62（S1）：63.
[107] AZZOPARDI L，GIROLAMI M，VAN RISJBERGEN K.Investigating the relationship between language model perplexity and IR precision-recall measures[C]//Proceedings of the 26th Annual International ACM SIGIR Conference on Research and Development in Informaion Retrieval，2003：369-370.
[108] PAPINENI K，ROUKOS S，WARD T，et al.Bleu：a method for automatic evaluation of machine translation[C]//Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics，2002：311-318.
[109] YU D，YAO K，SU H，et al.KL-divergence regularized deep neural network adaptation for improved large vocabulary speech recognition[C]//2013 IEEE International Conference on Acoustics，Speech and Signal Processing，2013：7893-7897.
[110] SONG Q，SUN Q，GUO D，et al.SinTra：learning an inspiration model from a single multi-track music segment[C]//International Society for Music Information Retrieval，2021：665-672.
[111] BRESIN R，FRIBERG A.Emotion rendering in music：range and characteristic values of seven musical variables[J].Cortex，2011，47（9）：1068-1081.
[112] MOGREN O.Continuous recurrent neural networks with adversarial training[J].arXiv：1611.09904，2016.
[113] CLARKE E F.Rhythm and timing in music[M]//The psychology of music.[S.l.]：Academic Press，1999：473-500.
[114] HONING H.9 structure and interpretation of rhythm in music[J].The Psychology of Music，2012：369.
[115] DONG H W，CHEN K，MCAULEY J，et al.MusPy：a toolkit for symbolic music generation[C]//International Society for Music Information Retrieval，2020：101-108.
[116] BURNS L.Analytic methodologies for rock music[J].Expression in Pop-Rock Music：Critical and Analytical Essays，2008：63-92.
[117] TYMOCZKO D.The geometry of musical chords[J].Science，2006，313（5783）：72-74.
[118] YEH Y C，HSIAO W Y，FUKAYAMA S，et al.Automatic melody harmonization with triad chords：a comparative study[J].Journal of New Music Research，2021，50（1）：37-51.
[119] UNEHARA M，ONISAWA T.Music composition system based on subjective evaluation[C]//2003 IEEE Intrnational Conference on Systems，Man and Cybernetics，2003：980-986.
[120] ZHU H，NIU Y，FU D，et al.MusicBERT：a self-supervised learning of music representation[C]//Proceedings of the 29th ACM International Conference on Multimedia，2021：3955-3963.
[121] WANG Z，XIA G.MuseBERT：pre-training music representation for music understanding and controllable generation[C]//International Society for Music Information Retrieval，2021：722-729.
[122] SONG K，TAN X，QIN T，et al.MASS：masked sequence to sequence pre-training for language generation[C]//International Conference on Machine Learning，2019：5926-5936.
[123] RAFFEL C，SHAZEER N，ROBERTS A，et al.Exploring the limits of transfer learning with a unified text-to-text transformer[J].Journal of Machine Learning Research，2020，21（140）：1-67.
[124] LEWIS M，LIU Y，GOYAL N，et al.BART：denoising sequence-to-sequence pre-training for natural language generation，translation，and comprehension[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics，2020：7871-7880.
[125] JEONG D，KWON T，KIM Y，et al.Graph neural network for music score data and modeling expressive piano performance[C]//International Conference on Machine Learning，2019：3060-3070.
[126] HAWTHORNE C，JAEGLE A，CANGEA C S，et al.General-purpose，long-context autoregressive modeling with perceiver AR[J].arXiv：2202.07765，2022.