Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
londons_explore
on Nov 25, 2016
|
parent
|
context
|
favorite
| on:
Speech-to-Text-WaveNet: End-to-end sentence level ...
Looking at the training loss graph, it looks like training for more time would produce even better results...
Anyone want to volunteer a few weeks of GPU time to train this better?
gwern
on Nov 26, 2016
[–]
Training loss pretty much always decreases. NNs are extremely powerful models, so they can overfit most data. What you want to see is the
validation
loss graph.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
Anyone want to volunteer a few weeks of GPU time to train this better?