Implicit Bias in Leaky ReLU Networks Trained on High-Dimensional Data
원문 초록
The implicit biases of gradient-based optimization algorithms are conjectured to be a major factor in the success of modern deep learning. In this work, we investigate the implicit bias of gradient flow and gradient descent in two-layer fully-connected neural networks with leaky ReLU activations when the training data are nearly-orthogonal, a common property of high-dimensional data. For gradient flow, we leverage recent work on the implicit bias for homogeneous neural networks to show that asymptotically, gradient flow produces a neural network with rank at most two. Moreover, this network is an $\ell_2$-max-margin solution (in parameter space), and has a linear decision boundary that corresponds to an approximate-max-margin linear predictor. For gradient descent, provided the random initialization variance is small enough, we show that a single step of gradient descent suffices to drastically reduce the rank of the network, and that the rank remains small throughout training. We provide experiments which suggest that a small initialization scale is important for finding low-rank neural networks with gradient descent.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Scaling Laws of Dynamic High-Capacity Ride-Sharing
CUTS+: High-dimensional Causal Discovery from Irregular Time-series
CosmoPower-JAX: high-dimensional Bayesian inference with differentiable cosmological emulators
HumanRF: High-Fidelity Neural Radiance Fields for Humans in Motion
MIReAD: Simple Method for Learning High-quality Representations from Scientific Documents
High-Fidelity 3D Face Generation from Natural Language Descriptions
토론
이 논문과 연결된 포럼 글이 아직 없습니다.