What can linear interpolation of neural network loss landscapes tell us?
This work addresses a methodological issue for researchers studying neural network optimization, showing that common visualization techniques may lead to incorrect inferences, which is incremental but clarifies a key point in the field.
The paper investigates whether linear interpolation of neural network loss landscapes reliably indicates optimization success, finding that the shape of linear paths does not correlate with changes in test accuracy, thus challenging prior assumptions.
Studying neural network loss landscapes provides insights into the nature of the underlying optimization problems. Unfortunately, loss landscapes are notoriously difficult to visualize in a human-comprehensible fashion. One common way to address this problem is to plot linear slices of the landscape, for example from the initial state of the network to the final state after optimization. On the basis of this analysis, prior work has drawn broader conclusions about the difficulty of the optimization problem. In this paper, we put inferences of this kind to the test, systematically evaluating how linear interpolation and final performance vary when altering the data, choice of initialization, and other optimizer and architecture design choices. Further, we use linear interpolation to study the role played by individual layers and substructures of the network. We find that certain layers are more sensitive to the choice of initialization, but that the shape of the linear path is not indicative of the changes in test accuracy of the model. Our results cast doubt on the broader intuition that the presence or absence of barriers when interpolating necessarily relates to the success of optimization.