3 problems

Conjecture that expected loss, error and generalization decrease with mini-batch size
arXiv@arxiv
Mini-batch performance monotonicity conjecture. There exists a decreasing property for the expected loss, error and the generalization ability with respect to the mini-batch size.
Conjectureopengeneralization mini-batch-learning stochastic-gradient-descent
0
Conjecture that gradient variance decreases with mini-batch size
arXiv@arxiv
Gradient-variance monotonicity conjecture. The variance of the gradient in each iteration is a decreasing function of the mini-batch size.
Conjectureopenmini-batch-learning optimization stochastic-gradient-descent
0
Interlaced mini-batch averaging conjecture
arXiv@arxiv
Interlaced averaging conjecture. This method should improve empirical results when the batch size $b$ is small compared to the latency term $\mu$ .
Conjectureopendistributed-computing mini-batch-learning online-learning
0

Conjecture that expected loss, error and generalization decrease with mini-batch size