Timeline for Overfitting in neural network
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 3, 2018 at 19:07 | comment | added | Akababa | Thanks, that was my thought too but I got confused by the diagram description. | |
| Jan 2, 2018 at 11:25 | comment | added | n1k31t4 | @Akababa: Reg your original question: the green line indicates little overfitting, I wouldn't say underfitting. A neural network is theoretically a universal approximator, meaning with enough neurons, it can map any input to any output i.e. it can approximate any function. Of course, if the model is underfitting, not performing well (and e.g. a lack of data is not the problem), results may be improved by inreasing the number of parameters, and so the flexibility of the model - allowing it to more closely approximate the function. Possible drawbacks: memory/training time/interpretability. | |
| Dec 27, 2017 at 15:17 | comment | added | Akababa | From the diagram: "The other possible case is when the validation accuracy tracks the training accuracy fairly well. This case indicates that your model capacity is not high enough: make the model larger by increasing the number of parameters." | |
| Dec 27, 2017 at 14:12 | comment | added | n1k31t4 | @Akababa - what do you mean? who mentioned underfitting? | |
| Dec 26, 2017 at 21:12 | comment | added | Akababa | Why does the green line indicate underfitting? | |
| Aug 12, 2017 at 11:18 | history | edited | n1k31t4 | CC BY-SA 3.0 |
added 8 characters in body
|
| Jul 21, 2017 at 13:53 | history | migrated | from stackoverflow.com (revisions) | ||
| Jul 20, 2017 at 21:55 | comment | added | n1k31t4 | Dropout is a good way to minimise such effects. There are several variants of it. To improve performance further, there are things like batch normalisation, layer normalisation and n other approaches - where n is large :) | |
| Jul 20, 2017 at 21:15 | comment | added | Ricky Han | Just use dropout. Your model will never overfit ever again. | |
| Jul 20, 2017 at 11:49 | vote | accept | CommunityBot | ||
| Jul 20, 2017 at 11:05 | comment | added | n1k31t4 | At each iteration, you can simply save the training and test accuracies, then plot them afterwards. Keras/Tensorflow also have objects that will track such information for you too, if you don't want to do it manually. It can be done with Tensorboard live during training (if you want to make things really cool)... Have a look at this: keras.io/callbacks/#tensorboard | |
| Jul 20, 2017 at 10:55 | comment | added | Damith Tilakaratne | as i know training data is used to train the model changing its weights and validation set is used to measure accuracy in trained model. so how can i plot a graph with training accuracy and validation accuracy? | |
| Jul 20, 2017 at 10:33 | history | answered | n1k31t4 | CC BY-SA 3.0 |