- Implement sequential fully connected neural network
- Implement sequential convolutional neural network
- Parallelize convlutional neural network with CUDA
- Give multiple variables derivative of 1 if they are within learning rate of each other if one of them is max
- Is there a way to learn how to learn? learn which features are better to learn? learn whether it's better to have 3x3 filters or 5x5 filters or 3x3 filters that then are used to devise 5x5 filters?
- Similarly is there a way to determine how to learn how to do linear regression? Don't have to search entire space of models, but try basic models and move in the right direction . . .
- Why wasn't mmap able to work on spiedie?