Showing cs.LGShow all
2 papers · 1 filter
cs.LG2025
The Butterfly Effect: Neural Network Training Trajectories Are Highly Sensitive to Initial Conditions
Devin Kwok, Gül Sena AltıntaÅ, Colin Raffel +1
Neural network training is inherently sensitive to initialization and the randomness induced by stochastic gradient descent. However, it is unclear to what extent such effects lead…
cs.LG2024
Simultaneous linear connectivity of neural networks modulo permutation
Ekansh Sharma, Devin Kwok, Tom Denton +3
Neural networks typically exhibit permutation symmetries which contribute to the non-convexity of the networks' loss landscapes, since linearly interpolating between two permuted v…