issues
search
barapa
/
lenin
Automatic chord transcription
3
stars
0
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
Write model analysis script for hessian free models
#31
barapa
opened
11 years ago
0
Save error rate in model with Hessian Free code
#30
barapa
opened
11 years ago
0
Use validation set to decide on best epoch and save that model
#29
barapa
opened
11 years ago
0
Normalize inputs
#28
barapa
closed
11 years ago
1
Combine left and right frames for the SVM data vectors.
#27
barapa
closed
11 years ago
1
Have one nn_error plot for all song batches.
#26
smessing
closed
11 years ago
0
Save nn error charts to a pdf file in /var/data/lenin/rbm_dbn_models (or somewhere else more appropriate)
#25
smessing
opened
11 years ago
0
Implement CD-N instead of CD-1
#24
barapa
closed
11 years ago
2
Write script to summarize SAE models
#23
barapa
closed
11 years ago
1
Make generate run scripts more general for different layers and SAE models
#22
barapa
closed
11 years ago
6
Use frames from the timestamps given in the chroma data
#21
barapa
opened
11 years ago
0
Combine consecutive frames for visible vector
#20
barapa
closed
11 years ago
1
Get Stacked Autoencoder Running
#19
barapa
closed
11 years ago
1
Write out reconstruction error and train / validation error during dbn training to log file.
#18
smessing
opened
11 years ago
0
Put more info in filename of svm data files to indicate which layers were included as features
#17
barapa
closed
11 years ago
0
Add logging to the NN training
#16
barapa
closed
11 years ago
1
Replace all mkdir() calls to ensure_dir_exists.m
#15
barapa
closed
11 years ago
2
Reduce size of input with different spectrogram settings.
#14
smessing
closed
11 years ago
3
Code to generate SVM training and testing scripts
#13
smessing
opened
11 years ago
0
Summarize models
#12
smessing
closed
11 years ago
0
Reorganize source folder
#11
smessing
closed
11 years ago
0
Don't sample on reconstruction, just use activation.
#10
smessing
closed
11 years ago
0
Song batches for training NN with backprop
#9
smessing
closed
11 years ago
1
Allow FFNN parameters to be fully customized through wrapper.
#8
barapa
closed
11 years ago
1
Fix minibatches so irregularly sized minibatch still gets run and scaled properly
#7
barapa
closed
11 years ago
0
Canonical Permutations
#6
barapa
closed
11 years ago
3
Gaussian inputs
#5
barapa
closed
11 years ago
4
ZCA Whitening
#4
barapa
closed
11 years ago
1
Increase momentum during training, and possibly decrease learning rate.
#3
barapa
opened
11 years ago
0
Adjust gradient wrt the minibatch size
#2
barapa
closed
11 years ago
4
Probabilities vs stochastic binary states
#1
barapa
closed
11 years ago
0