input sequence for lstm network

1 次查看（过去 30 天）
Philip Hua
Philip Hua 2022-6-3
回答： Philip Hua 2022-6-5
hi - i am sorry for asking many questions on LSTM but i find the implementation very confusing in matlab. I have a 32x305296 array of doubles which is generated by embedding musical note tokens in 32 dimensions. Effectively, we have 305296 notes of dimension 32 each in a time sequence. I need to create a LSTM network to train and predict the next note. (so it's a classification problem rather than regression as the token must be one of the permissible notes). I have set up the NN below. When i run the code I get this error
"Invalid training data. For image, sequence-to-label, and feature classification tasks, responses must be categorical."
Bearing in mind that this is a time sequence of embedded notes and not a group of labels, how do i change the time series of tokens to a permissible type?
layers = BachBotNN(32,32,0.3);
function [layers]=BachBotNN(inputSize,numHiddenUnits,dropoutProb)
layers = [ ...
sequenceInputLayer(inputSize)
lstmLayer(numHiddenUnits,'OutputMode','sequence')
batchNormalizationLayer
dropoutLayer(dropoutProb)
lstmLayer(numHiddenUnits,'OutputMode','sequence')
batchNormalizationLayer
dropoutLayer(dropoutProb)
lstmLayer(numHiddenUnits,'OutputMode','last')
batchNormalizationLayer
dropoutLayer(dropoutProb)
fullyConnectedLayer(1)
softmaxLayer
classificationLayer
]
  1 个评论
Philip Hua
Philip Hua 2022-6-5
Don't worry about this. I have managed to figure it out using an example online. Thank you

请先登录，再进行评论。

请先登录，再回答此问题。

采纳的回答

Philip Hua
Philip Hua 2022-6-5
there is an example of how to do this in scripttunegenerator.mlx. The whole process is not intuitive so it does take a lot of time to figure out something trivial.

更多回答（0 个）

类别

在 Help Center 和 File Exchange 中查找有关 Sequence and Numeric Feature Data Workflows 的更多信息

标签

产品


版本

R2021b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by