ICASSP 2019音声&音響論文読み会(https://connpass.com/event/128527/)での発表資料です。
AASP (Audio and Acoustic Signal Processing) 分野の紹介と、ICASSP 2019での動向を紹介しています。#icassp2019jp
Beyond the Symbols: A 30-minute Overview of NLPMENGSAYLOEM1
This presentation delves into the world of Natural Language Processing (NLP), exploring its goal to make human language understandable to machines. The complexities of language, such as ambiguity and complex structures, are highlighted as major challenges. The talk underscores the evolution of NLP through deep learning methodologies, leading to a new era defined by large-scale language models. However, obstacles like low-resource languages and ethical issues including bias and hallucination are acknowledged as enduring challenges in the field. Overall, the presentation provides a condensed, yet comprehensive view of NLP's accomplishments and ongoing hurdles.
The ArangoML Group had a detailed discussion on the topic "GraphSage Vs PinSage" where they shared their thoughts on the difference between the working principles of two popular Graph ML algorithms. The following slidedeck is an accumulation of their thoughts about the comparison between the two algorithms.
Presentation slide for AI seminar at Artificial Intelligence Research Center, The National Institute of Advanced Industrial Science and Technology, Japan.
URL (in Japanese): https://www.airc.aist.go.jp/seminar_detail/seminar_046.html
ICASSP 2019音声&音響論文読み会(https://connpass.com/event/128527/)での発表資料です。
AASP (Audio and Acoustic Signal Processing) 分野の紹介と、ICASSP 2019での動向を紹介しています。#icassp2019jp
Beyond the Symbols: A 30-minute Overview of NLPMENGSAYLOEM1
This presentation delves into the world of Natural Language Processing (NLP), exploring its goal to make human language understandable to machines. The complexities of language, such as ambiguity and complex structures, are highlighted as major challenges. The talk underscores the evolution of NLP through deep learning methodologies, leading to a new era defined by large-scale language models. However, obstacles like low-resource languages and ethical issues including bias and hallucination are acknowledged as enduring challenges in the field. Overall, the presentation provides a condensed, yet comprehensive view of NLP's accomplishments and ongoing hurdles.
The ArangoML Group had a detailed discussion on the topic "GraphSage Vs PinSage" where they shared their thoughts on the difference between the working principles of two popular Graph ML algorithms. The following slidedeck is an accumulation of their thoughts about the comparison between the two algorithms.
Presentation slide for AI seminar at Artificial Intelligence Research Center, The National Institute of Advanced Industrial Science and Technology, Japan.
URL (in Japanese): https://www.airc.aist.go.jp/seminar_detail/seminar_046.html
2. Lecture Part1 Introduction to Music Generation
♪ Today’s Goal
♪ 音楽生成、その挑戦の過程
♪音楽生成の歴史
♪生成方式の類型化
♪さらなる発展へ向けて
♪ Google Magentaとは
Agenda
3. Lecture Part2 Dive into Google Magenta
♪ Google Magentaの仕組み
♪Magentaにとっての「音」と「音楽生成」
♪Neural Network: シンプルなモデルによる、「音」の予測
♪RNN: 「メロディー」をとらえた生成への挑戦
♪LSTM: より長いメロディーの理解への挑戦
♪Lookback: 「リピート」への着目
♪Attention: 「キーポイント」をとらえたメロディーの生成へ
Agenda
4. Hands-on Part
♪ Hands-on Goal
♪ Magentaで、MIDIデータを学習させて音楽を生成する
♪ Magentaで作成したモデルと共演して、音楽を制作する
Hands-on document is available here.
At the End
♪ Toward the Next Music Production
Agenda
30. Google Magentaとは(2/2)
Google Magentaを利用するメリット
さらなる研究のスタートポイントとして
Google Magentaを利用することで、先に上げた研究の入り口まですぐにた
どり着くことができます。また、その先のモデルを構築する際に付属の仕
組み(音楽データの読み取りなど)は有用です(だがPython2)。
さらなる活用のスタートポイントとして
学習済みのモデルが提供されているため、あまり難しいことが分からなく
ても研究成果を利用すること可能です。実際、AI Duetのような動かすこと
のできるアプリケーションも公開されてます。
Let’s Go with Magenta!
57. 参考文献(1/2)
Resources
♪ Wikipedia: 自動作曲
♪ モーツァルト 音楽のサイコロ遊び
♪ 音楽における自動処理と Directability
♪ 《イリアック組曲》と『実験音楽』 コンピュータ音楽の創作を対象とした研究の一事例として
♪ セリー音楽 | 現代美術用語辞典ver.2.0 - Artscape
♪ 役に立つ音楽の情報~専門学校 十二音技法
♪ AIはプリペアド・ピアノの夢を見るか?――人工知能と自動作曲に関する覚書
♪ 魔法のアプリ、Chordana Composerが持つ作曲テクニックを探る
♪ 話題の人工知能作曲システムAmper Musicはまともな作曲システムなのか?
♪ ループシーケンス機能でオリジナル曲を作ってみよう(1)
♪ Magenta:人工知能と作曲
♪ ANALYZING SIX DEEP LEARNING TOOLS FOR MUSIC GENERATION
♪ Long short-term memory
♪ なぜミュージシャンはローランド創業者の訃報を受けて、悲しみに暮れたのか
♪ create with.AI
Training Course
♪ Machine Learning
♪ The Technology of Music Production
♪ Audio Signal Processing for Music Applications
58. 参考文献(2/2)
Products
♪ CASIO Chordana Composer
♪ Amper Music
♪ Neutron
Projects
♪ Magenta
♪ A.I. Duet
♪ SongFrom PI
♪ FLOWCOMPOSER: COMPOSING WITH AI
♪ AudioSet
Articles
♪ SampleRNN: An Unconditional End-to-End Neural Audio Generation Model
♪ Structured Attention Networks
♪ TUNING RECURRENT NEURAL NETWORKS WITH REINFORCEMENT LEARNING
♪ MidiNet: A Convolutional Generative Adversarial Network for Symbolic-domain Music Generatio
n using 1D and 2D Conditions
♪ Song From PI: A Musically Plausible Network for Pop Music Generation
♪ Neural Audio Synthesis of Musical Notes with WaveNet Autoencoders
♪ WaveNet: A Generative Model for Raw Audio