ANODE: Unconditionally Accurate Memory-Efficient Gradients for Neural ODEs
arXiv:1902.10298
Abstract
Residual neural networks can be viewed as the forward Euler discretization of an Ordinary Differential Equation (ODE) with a unit time step. This has recently motivated researchers to explore other discretization approaches and train ODE based networks. However, an important challenge of neural ODEs is their prohibitive memory cost during gradient backpropogation. Recently a method proposed in [8], claimed that this memory overhead can be reduced from O(LN_t), where N_t is the number of time steps, down to O(L) by solving forward ODE backwards in time, where L is the depth of the network. However, we will show that this approach may lead to several problems: (i) it may be numerically unstable for ReLU/non-ReLU activations and general convolution operators, and (ii) the proposed optimize-then-discretize approach may lead to divergent training due to inconsistent gradients for small time step sizes. We discuss the underlying problems, and to address them we propose ANODE, an Adjoint based Neural ODE framework which avoids the numerical instability related problems noted above, and provides unconditionally accurate gradients. ANODE has a memory footprint of O(L) + O(N_t), with the same computational cost as reversing ODE solve. We furthermore, discuss a memory efficient algorithm which can further reduce this footprint with a trade-off of additional computational cost. We show results on Cifar-10/100 datasets using ResNet and SqueezeNext neural networks.
References in corpus (5)
Cited by in corpus (39)
- Stiff Neural Ordinary Differential Equations
- Neural Controlled Differential Equations for Irregular Time Series
- Equivariant Flows: Exact Likelihood Generative Learning for Symmetric Densities
- Graph Neural Ordinary Differential Equations
- Learning Long-Term Dependencies in Irregularly-Sampled Time Series
- Signatory: differentiable computations of the signature and logsignature transforms, on both CPU and GPU
- Adaptive Checkpoint Adjoint Method for Gradient Estimation in Neural ODE
- Galaxy Morphology Classification using Neural Ordinary Differential Equations
- Discretize-Optimize vs. Optimize-Discretize for Time-Series Regression and Continuous Normalizing Flows
- ANODEV2: A Coupled Neural ODE Evolution Framework
- Continuous-in-Depth Neural Networks
- Causal Navigation by Continuous-time Neural Networks
- Approximation Capabilities of Neural ODEs and Invertible Residual Networks
- Learning Neural Event Functions for Ordinary Differential Equations
- NeuPDE: Neural Network Based Ordinary and Partial Differential Equations for Modeling Time-Dependent Data
- SNODE: Spectral Discretization of Neural ODEs for System Identification
- A Neural Network Approach Applied to Multi-Agent Optimal Control
- Momentum Residual Neural Networks
- Interpolation Technique to Speed Up Gradients Propagation in Neural ODEs
- Machine learning structure preserving brackets for forecasting irreversible processes
- Real-time Classification from Short Event-Camera Streams using Input-filtering Neural ODEs
- Stochasticity in Neural ODEs: An Empirical Study
- Inverse modified differential equations for discovery of dynamics
- Neural Ordinary Differential Equations for Data-Driven Reduced Order Modeling of Environmental Hydrodynamics
- MRI Image Reconstruction via Learning Optimization Using Neural ODEs
- Towards Understanding Normalization in Neural ODEs
- Opening the Blackbox: Accelerating Neural Differential Equations by Regularizing Internal Solver Heuristics
- Spline parameterization of neural network controls for deep learning
- Neural ODEs for Image Segmentation with Level Sets
- Quantized Convolutional Neural Networks Through the Lens of Partial Differential Equations
- ODE-based Deep Network for MRI Reconstruction
- Learning Runge-Kutta Integration Schemes for ODE Simulation and Identification
- Faster Policy Learning with Continuous-Time Gradients
- A Practical Layer-Parallel Training Algorithm for Residual Networks
- Second-Order Neural ODE Optimizer
- Meta-Solver for Neural Ordinary Differential Equations
- Learning ODEs via Diffeomorphisms for Fast and Robust Integration
- Delay Differential Neural Networks
- Incorporating NODE with Pre-trained Neural Differential Operator for Learning Dynamics