You get access to the resulting Keras model, so you can iterate on it manually. Great way to establish a solid baseline on a new problem (as long as your dataset is small enough for architecture search to be tractable).
Show this thread
Did you know you could use AutoKeras to quickly develop simple image & text models without having to make any configuration choices yourself? It will do architecture search and hyperparameter tuning for you in the background. https://autokeras.com/ pic.twitter.com/3d644bgucK
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Any reports on the accuracies and how the configuration search works?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Do you think this could be used on a feature vector input instead of an image? Because then you could use to make the transfer learning part more automatic. Does I'm makes sense?
-
If you already have feature vectors, you could try StructuredDataClassifier
End of conversation
New conversation -
-
-
I shall try this out
End of conversation
New conversation -
-
Very excited to try this (manually coded a simile function). But this simple approach throws a “DirectoryIterator” object has no shape” error. I’m guessing it struggles with ImageDataGenerator.
-
Use http://tf.data datasets or just numpy arrays
- Show replies
New conversation -
-
-
Data science is becoming drag&drop .. only problem is if you give kinfe to a monkey ... it will be a disaster.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Amazing amazing stuff.
@fchollet, you rock
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.