Re:A Neural Network in 11 Lines of Python (2015)
by mholt on Wednesday October 18, 2017 @05:07PM
(#15503109)
I co-authored a blog post with my lab that has practical advice for debugging DNNs - some of these tips might be helpful to you? https://pcc.cs.byu.edu/2017/10/02/practical-advice-for-build...
Re:A Neural Network in 11 Lines of Python (2015)
by maurits on Thursday October 19, 2017 @04:11AM
(#15506241)
Stanford CS231n: Convolutional Neural Networks for Visual Recognition [1]
The assignments are excellent and will let you implement a deephish network from practically scratch, including backprop, optimizers, tuning etc.
Re:A Neural Network in 11 Lines of Python (2015)
by jackthetab on Wednesday October 18, 2017 @07:37PM
(#15504250)
You may want to check out Andrew Ng's Deep Learning Specialization over on Coursera. [1] One of the courses is specifically about hyperparameter tuning and another about structuring your project. There is a lot of practical information scattered across all the courses.
Yes, I'm taking the specialization and having a blast with it. :-)
[1] https://www.coursera.org/learn/neural-networks-deep-learning
Re:A Neural Network in 11 Lines of Python (2015)
by T_D_K on Wednesday October 18, 2017 @04:53PM
(#15502947)
Based on the limited amount of information, I'm assuming that by "training explodes" you mean that your gradient descent never reaches a local minimum. Try lowering your learning rate? You may be "stepping over" the minimum.
It sounds like there is a growing bag of tricks neural network researchers are discovering to make training practical and stable for large data sets.
One example would be using relu activation - whenever I play with it in a simple tutorial like this one, training seems to explode and fail much more frequently, so I'm guessing either I'm missing another step people use, or there are some extra constraints on initial conditions?
Using a Gaussian for activation in my tutorials has tended to be more stable and converge much faster, but I assume there is a huge downside lurking somewhere to having a non-monotonically increasing function?
What are the tricks of the trade that a weekend warrior should investigate?