# VAE loss doesn't converge to zero. Does it make sense to sample new instances from trained latent space?

I aim to use a variational autoencoder (VAE) as a generative model. Does this make sense only if the reconstruction loss converges towards zero?

On a project I'm working on, the loss is getting reduced but at some point it won't converge to 0.

Epoch 1 of 1000
100%|██████████| 506/506 [00:07<00:00, 63.36it/s]
Train Loss: 31.2318
Epoch 2 of 1000
100%|██████████| 506/506 [00:07<00:00, 63.57it/s]
Train Loss: 19.9676
Epoch 3 of 1000
100%|██████████| 506/506 [00:08<00:00, 61.88it/s]
Train Loss: 19.0511
Epoch 4 of 1000
100%|██████████| 506/506 [00:08<00:00, 63.07it/s]
Train Loss: 18.5793
Epoch 5 of 1000
100%|██████████| 506/506 [00:08<00:00, 62.91it/s]
Train Loss: 18.1751


And it won't improve that much.

So the network is learning, maybe just not that much as it should. Despite this, does it still make sense to sample new instances from the trained latent space?

• You might also want to output the parameters of the likelihood distribution $p(x|z)$ with the decoder. What I mean by that is that depending on the type of data it should output parameters of a distribution (similar to the encoder), parameters of a normal distribution for continuous data and parameters for a bernoulli distribution for binary data. Then you can use the negative log likelihood as loss function. This change helps the model to better express uncertainty. May 5 at 18:45