Barcodes as summary of objective function's topology
We apply the canonical forms (barcodes) of gradient Morse complexes to explore topology of loss surfaces. We present a novel algorithm for calculations of the objective function's barcodes of local minima. We have conducted experiments for calculating barcodes of local minima for benchmark functions and for loss surfaces of neural networks. Our experiments confirm two principal observations for loss surfaces of neural networks. First, the barcodes of local minima are located in a small lower part of the range of values of loss function of neural networks. Second, increase of the neural network's depth brings down the barcodes of local minima. This has natural implications for the neural network learning and the generalization ability.
PDF Abstract