photo

J


Last seen: 21 jours il y a |  Actif depuis 2018

Followers: 0   Following: 0

Message

Statistiques

  • Knowledgeable Level 1
  • First Answer
  • Revival Level 1

Afficher les badges

Feeds

Afficher par

Réponse apportée
Cross-Entropy Loss in Classification Layers
I ended up just writing my own custom layers. Another language is an option, but I also sympathize with those who have a large ...

environ 5 ans il y a | 1

Question


Neural Network Checkpoints only every 'N' Epochs??
In the MATLAB documentation for Deep Learning Toolbox (formerly Neural Network Toolbox) we see that we can use <https://www.math...

presque 6 ans il y a | 1 réponse | 1

0

réponse

Question


Cross-Entropy Loss in Classification Layers
The MATLAB documentation is very clear to say that a <https://www.mathworks.com/help/deeplearning/ref/classificationlayer.html c...

presque 6 ans il y a | 1 réponse | 0

1

réponse

Réponse apportée
R2018a Neural Network Toolbox "Bugs"??
At least for (2), the issue is that checkpoints don't save the state of the solver (momentum parameters, etc), so your solver ca...

plus de 6 ans il y a | 1

| A accepté

Réponse apportée
How to manually change pre-trained CNN's (AlexNet) weights?
Bump! I have a hard time believing that one cannot take an existing SeriesNetwork, and either: 1. Modify properties of it, o...

plus de 6 ans il y a | 1

Réponse apportée
R2018a Neural Network Toolbox "Bugs"??
Here is an example of what the validation plot looks like at the very end with the consistent "drop" in the "Final" data points'...

plus de 6 ans il y a | 0

Question


R2018a Neural Network Toolbox "Bugs"??
In R2018a Neural Network Toolbox I have noticed two oddities: (1) When training a CNN with trainNetwork and 'options' field ...

plus de 6 ans il y a | 3 réponses | 0

3

réponses

Réponse apportée
why using imageDataAugmenter doesn't increase the size of my training data set ?
I am guessing that when augmentation is on, it trains "exactly" the way it trained when it was turned off, but performs a random...

plus de 6 ans il y a | 1

| A accepté