To learn languages based on projects. Github: https://github.com/tuvtran/project-based-learning
Python Machine Learning Book Github: https://github.com/rasbt/python-machine-learning-book
Coding Practice and Algorithms Github: https://github.com/jwasham/coding-interview-university
What every programmer should know Github: https://github.com/mtdvio/every-programmer-should-know
Awesome public datasets Github: https://github.com/awesomedata/awesome-public-datasets
Awesome Machine Learning Github: https://github.com/josephmisiti/awesome-machine-learning
Awesome Deep Vision Github: https://github.com/kjw0612/awesome-deep-vision
Awesome tensorflow Github: https://github.com/jtoy/awesome-tensorflow
Awesome Project Ideas Github: https://github.com/NirantK/awesome-project-ideas
Awesome NLP Github: https://github.com/keon/awesome-nlp
Best of Jupyter Github: https://github.com/NirantK/best-of-jupyter
Deep Learning paper reading roadmap Github: https://github.com/floodsung/Deep-Learning-Papers-Reading-Roadmap
Paper to Code Github: https://github.com/zziz/pwc
Reinforcement Learning Github: https://github.com/dennybritz/reinforcement-learning
Google dataset search Link: https://t.co/iXFwNCDaUN
Best Practices for ML Engineering Link: http://martin.zinkevich.org/rules_of_ml/rules_of_ml.pdf
List of Tutorials - Medium Article Link: https://medium.com/machine-learning-in-practice/over-200-of-the-best-machine-learning-nlp-and-python-tutorials-2018-edition-dd8cf53cb7dc
Awesome list of people and blogs to follow to keep yourself updated in the field Link: https://medium.com/@alexrachnog/ultimate-following-list-to-keep-updated-in-artificial-intelligence-32776ffcd079
Google's guide to Machine Learning Link: https://techdevguide.withgoogle.com/paths/machine-learning/
Nothing makes you feel stupid quite like reading a scientific journal article.
- How to read a scientific paper
- Getting started with reading Deep Learning Research papers: The Why and the How
- Deep Learning Techniques for Music Generation a Survey
- Machine-Learning for Symbolic Music Generation
- Dicas para aprender Machine Learning
- Reconhecimento de escrita manual com Redes Neurais Convolucionais
- Aprendizado de Máquina
- http://www.dobitaobyte.com.br/primeiros-passos-com-tensorflow-e-minhas-anotacoes/
- Redes Neurais Artificiais
- Como visualizar um modelo de rede neural de aprendizado profundo em Keras
- Redes Neurais com Keras
- APLICAÇÃO DE REDES NEURAIS GERADORAS ADVERSÁRIAS PARA COLORIZAÇÃO DE IMAGENS EM PRETO E BRANCO
Montem uma GAN usando LSTM mas por enquanto nao funcionou. Baseado nesta
Teste GPU
from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())Verificar GPU
$ nvcc --version
$ nvidia-smiproblemas com a gpu:
ImportError: libcublas.so.9.0: cannot open shared object file: No such file or directory
isso é porque o tensorflow 1.5 acima não funciona no CUDA 8.0. Para resolver isso instale a verso 1.4
$ pip install --upgrade tensorflow-gpu==1.4Download all files in a path on Jupyter notebook server
DeepJazz Coursera instalar módulo qa
instalar também grammar
pip install anovelmous_grammar
Python online sem precisar de instalar - repl.it
Deep Learning for Music (DL4M)
us = environment.UserSettings()
us.getSettingsPath()
'/home/usuario/.music21rc'
us['musicxmlPath']='/usr/bin/musescore'
us['musescoreDirectPNGPath']='/usr/bin/musescore'
Vários artigos - Music Generation With DL
Excelentes materiais de estudos de Machine Learning
Resolver o problema do Kernel não aparecer no jupyter notebook
source activate myenv
python -m ipykernel install --user --name myenv --display-name "Python (myenv)"
source activate other-env
python -m ipykernel install --user --name other-env --display-name "Python (other-env)"
Text Generation using Generative Adversarial Networks (GAN) - Core challenges
Depois de ter instalado o Cuda 9 (errado porque o tensorflow ainda nao funciona com essa versao). Deu muito trabalho para conseguir remover esta e instalar o Cuda 8. O que funcionou foi a dica desse site.
NETWORK FOR SYMBOLIC-DOMAIN MUSIC GENERATION](https://arxiv.org/pdf/1703.10847.pdf) Código
Para rodar o projeto https://github.com/llSourcell/Music_Generator_Demo além do tensorflow deve instalar tdm e o python-midi
$ pip install https://pypi.python.org/packages/71/3c/341b4fa23cb3abc335207dba057c790f3bb329f6757e1fcd5d347bcf8308/tqdm-4.19.5-py2.py3-none-any.whl#md5=99fe93af82488226c1afadfaec0e1498
$ # testei no python3 e deu certo
$ git clone https://github.com/louisabraham/python3-midi.git
$ #dentro do diretório clonado rodar o comando
$ python setup.py install
$ import os
$ import sys
$ os.path.dirname(sys.executable)
- MIDI NET
- MUSEGAN: DEMONSTRATION OF A CONVOLUTIONAL GAN BASED MODEL FOR GENERATING MULTI-TRACK PIANO-ROLLS
- Tutorial Pretty-Midi
-
Para fazer rodar o código https://github.com/olofmogren/c-rnn-gan só com o python 2.7 e tensorflow 0.12.1 e python-midi
$ conda create -n rnn-gan python=2.7 $ source activate rnn_gan $ export TF_BINARY_URL=https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.12.1-cp27-none-linux_x86_64.whl $ pip install --upgrade $TF_BINARY_URL $ pip install python-midi $ python rnn_gan.py --datadir datadir --traindir traindir -testdir test
Leitura do artigo Algorithmic Composition of Melodies with Deep Recurrent Neural Networks
Variational Autoencoder: Intuition and Implementation in Keras
- Vanilla LSTM with numpy
- Generating Random Vectors from the Multivariate Normal Distribution
- Normal Distribution
- (Effective TensorFlow)Machine Learning Top 10 Articles For the Past Month (v.Sep 2017)
- Creating A Language Translation Model Using Sequence To Sequence Learning Approach
- Matemática para Deep Learning
- Algebra Linear Livros(http://www.uesb.br/professor/flaulles/download/softwares/IntAlgebraLinear.pdf)
- Deep Learning Mindmap / Cheatsheet
- Encoder-Decoder Long Short-Term Memory Networks
- Music Generation Using Rnn
- Delving deep into Generative Adversarial Networks (GANs)
- https://pythonprogramming.net/machine-learning-tutorial-python-introduction/
- How to Use an Encoder Decoder Lstm to Echo Sequences of Random Integers
- Tutorials Keras Recurrent
- Python Cheat Sheet Basic
- Python Cheat Sheet Intermediate
- Pandas Cheat Sheet
- Pandas Tutorial Python 2
- Naive Bayes Tutorial
- Vários links de projetos de música usando deep learning
- Leitura do artigo Song From PI A Musically Plausible Network for Pop Music Generation
- Programming instrumental music from scratch - Ensina o funcionamento da biblioteca Python MIDI
- Music Note Symbols
- Deep Learning From Scratch: Theory and Implementation
- DEEP LEARNING METALLICA WITH RECURRENT NEURAL NETWORKS
- Isolated guitar transcription using a deep belief network
-
Configurar o Music21 para usar o Musescore no Windows:
$ us = environment.UserSettings() $ us['musicxmlPath']="C:\\Program Files (x86)\\MuseScore 2\\bin\\MuseScore.exe" $ us["musescoreDirectPNGPath"]="C:\\Program Files (x86)\\MuseScore 2\\bin\\MuseScore.exe"
- https://henri.io/posts/chord-classification-using-neural-networks.html
- https://www.thisismetis.com/made-at-metis
- Geração de música JohaNN após treinar a rede
- The 5 Step Life-Cycle for Long Short-Term Memory Models in Keras
- Generative Adversarial Networks Part 2 - Implementation with Keras 2.0
- https://www.analyticsvidhya.com/blog/2017/06/introductory-generative-adversarial-networks-gans/
- Estudo do código
- Exemplo com music21 mid_online_music21
-
Estudos da biblioteca Music21 e Pyknon baseados no tutorial
-
Consegui fazer funcionar o áudio no Musescore instalando o gawk:
$ sudo apt-get install gawk
- Estudando como transformar MIDI em rolo de Piano. midiToNoteStateMatrix. Não quer funcionar para arquivos MIDI para violão. Mas para os datasets do projeto de exemplo funciona. É interessante a representaço onde a metade da matriz contém os bits um's para indicar quando a nota foi tocada.
- Uma forma interessante de representar um rolo de piano
- Visualização de rolo de piano em music21
- Estudo da biblioteca pretty_midi Paper INTUITIVE ANALYSIS, CREATION AND MANIPULATION OF MIDI DATA WITH pretty_midi
- Estudo do código GAN-Sandbox em Keras/Tensorflow
- Leitura do artigo Text Generation using Generative Adversarial Training
- Estudo do código TextGAN
- Pesquisa sobre as bibliotecas music21, unroll e lilypond
- Leitura sobre transposição para ser utilizada em DeepLearning
- Leitura - Metis Final Project Music Composition
- Leitura parcial do projeto musicNet
-
Estudo do código sequence_gan
-
Estudo do código Modelling-and-Generating-Sequences-of-Polyphonic-Music-With-RNN-RBM
-
Estudo do códigoMusic Generator Demo. É necessário instalar os pacotes no python2:
$ pip install Numpy==1.11.0 $ pip install tensorflow==0.12.0 $ pip install pandas==0.19.2 $ pip install msgpack-python==0.4.8 $ pip install glob2==0.5 $ pip install tqdm==4.11.2 $ pip install python-midi==0.2.4 -
Estudo do código How-to-Generate-Music-Demo Para funcionar tem que remover a parte fixa mode='major' do arquivo preprocess.py Video. Se gerar com mais de 25 épocas dá erro o qual pode ser corrigido conforme essa dica. É o mesmo código de DeepJazz
-
Leitura Aprendizado por reforço
-
Estudo Set em Python
-
Leitura do artigo Deep Learning for Music
-
Leitura do artigo Objective-Reinforced Generative Adversarial Networks (ORGAN) for Sequence Generation Models Código
$ conda create -c rdkit -n organ rdkit python=2.7 $ source activate organ $ pip install tqdm -
Estudo de função de verossimilhança Código
-
Debug GAN
-
Debug SeqGAN
-
Estudo Tensorflow Getting Started With TensorFlow e MNIST For ML Beginners
- Leitura do projeto final Deep-Rock https://cs224d.stanford.edu/reports/Ilan.pdf
- Teste do Melody_RNN do projeto magenta: https://github.com/tensorflow/magenta/blob/master/magenta/models/melody_rnn/README.md Para isso foi criado as variáveis de ambiente abaixo e depois seguiu-se o tutorial.
$ docker pull tensorflow/magenta #download o docker do projeto magenta
$ docker run -it tensorflow/magenta bash
$ export BUNDLE_PATH=/magenta-models/basic_rnn.mag
$ export CONFIG=basic_rnn
$ melody_rnn_generate \
--config=${CONFIG} \
--bundle_file=${BUNDLE_PATH} \
--output_dir=/tmp/melody_rnn/generated \
--num_outputs=10 \
--num_steps=128 \
--primer_melody="[60]"
- Leitura https://magenta.tensorflow.org/2016/07/15/lookback-rnn-attention-rnn/
- Links docker http://stefanteixeira.com.br/2015/03/17/comandos-essenciais-docker-monitoramento-containers/ https://woliveiras.com.br/posts/imagem-docker-ou-um-container-docker/ Obs. Para sair do container sem matar a sessão digitar Ctrl+P+Q. Para entrar novamente digirar docker attach + containerId
- blog https://medium.com/towards-data-science/can-a-deep-neural-network-compose-music-f89b6ba4978d
- poster https://www.justinsvegliato.com/pdf/deep-jammer-poster.pdf
- artigo https://www.justinsvegliato.com/pdf/deep-jammer-report.pdf
- código https://github.com/justinsvegliato/deep-jammer
- Código https://github.com/hexahedria/biaxial-rnn-music-composition
- Blog http://www.hexahedria.com/2015/08/03/composing-music-with-recurrent-neural-networks/
- Artigo http://www.hexahedria.com/files/2017generatingpolyphonic.pdf
- Leituras relacionadas:RNN Music Composer, Creativity in Machine Learning, Composing Classical Music with Recurrent Neural Network
$ sudo apt-get install libasound2
$ sudo apt-get install libasound2-dev
$ pip install python-rtmidi
$ pip install magenta
$ conda create -n deep-improvisation python=2.7
$ conda-env remove -n deep-improvisation
$ source activate deep-improvisation
- Maxima verossimilhança
- GAN
| Titulo | Description | Site | Biblioteca |
|---|---|---|---|
| LSTM network for algorithmic music composition | Este é o repositório de código para o trabalho de tese de mestrado, em Composição de música algorítmica utilizando redes neuronais recorrentes (RNN) | https://github.com/bernatfp/LSTM-Composer | Keras e mido |
| LSTM-RNN-Melody-Composer | Esta é a implementação de uma Rede Neural Recorrente LSTM que compõe uma melodia para uma determinada sequência de acordes. | https://github.com/konstilackner/LSTM-RNN-Melody-Composer | Keras e Mido |
| JohaNN | 2-layer LSTM music generator trained on Bach Cello Suites | https://github.com/naoyak/JohaNN | Keras e music21 |
| SCHUMANN: RECURRENT NEURAL NETWORKS COMPOSING MUSIC | http://inspiratron.org/blog/2017/01/01/schumann-rnn-composing-music/ | ||
| Music Language Modeling with Recurrent Neural Networks | (https://github.com/yoavz/music_rnn) http://yoavz.com/music_rnn/ | ||
| StyleNet | Neural Translation of Musical Style | Dissertação Blog Código | Tensorflow,mido,pretty_midi,fluidsynth |
| Metis Final Project: Music Composition with LSTMs | a recurrent neural network utilizing Long Short-Term Memory nodes (LSTMs) to learn patterns in the Six Cello Suites by J.S. Bach | Blog | |
| neuralnetmusic | The myparser.py module turns musicxml files into training data | https://github.com/fephsun/neuralnetmusic | Keras |
| Music Generator Demo - Siraj Raval | Overview Use TensorFlow to generate short sequences of music with a Restricted Boltzmann Machine. This is the code for Generate Music in TensorFlow on YouTube. | https://github.com/llSourcell/Music_Generator_Demo | Python midi e Tensorflow |
| Music Generator Demo - Siraj Raval | This is the code for this video on Youtube by Siraj Raval as part of the the Udacity Deep Learning Nanodegree. It uses Keras & Theano, two deep learning libraries, to generate jazz music. Specifically, it builds a two-layer LSTM, learning from the given MIDI file. | https://github.com/llSourcell/How-to-Generate-Music-Demo | Keras, Theano, Music21 |
| LSTM-RNN-Melody-Composer | Esta é a implementação de uma Rede Neural Recorrente LSTM que compõe uma melodia para uma determinada sequência de acordes | https://github.com/konstilackner/LSTM-RNN-Melody-Composer | mido e Keras |
| Implementation of C-RNN-GAN. | Aplicado a geração de música | https://github.com/olofmogren/c-rnn-gan | Tensorflow |