Differentiable Convex Optimization Layers

Abstract

Recent work has shown how to embed differentiable optimization problems (that is, problems whose solutions can be backpropagated through) as layers within deep learning architectures. This method provides a useful inductive bias for certain problems, but existing software for differentiable optimization layers is rigid and difficult to apply to new settings. In this paper, we propose an approach to differentiating through disciplined convex programs, a subclass of convex optimization problems used by domain-specific languages (DSLs) for convex optimization. We introduce disciplined parametrized programming, a subset of disciplined convex programming, and we show that every disciplined parametrized program can be represented as the composition of an affine map from parameters to problem data, a solver, and an affine map from the solver's solution to a solution of the original problem (a new form we refer to as affine-solver-affine form). We then demonstrate how to efficiently differentiate through each of these components, allowing for end-to-end analytical differentiation through the entire convex program. We implement our methodology in version 1.1 of CVXPY, a popular Python-embedded DSL for convex optimization, and additionally implement differentiable layers for disciplined convex programs in PyTorch and TensorFlow 2.0. Our implementation significantly lowers the barrier to using convex optimization problems in differentiable programs. We present applications in linear machine learning models and in stochastic control, and we show that our layer is competitive (in execution time) compared to specialized differentiable solvers from past work.

In NeurIPS 2019. Code available at https://www.github.com/cvxgrp/cvxpylayers. Authors in alphabetical order

Cited by in corpus (66)

• Solving Inverse Problems With Deep Neural Networks -- Robustness Included?
• RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
• DC3: A learning method for optimization with hard constraints
• DeepEMD: Differentiable Earth Mover's Distance for Few-Shot Learning
• learn2learn: A Library for Meta-Learning Research
• Enforcing Policy Feasibility Constraints through Differentiable Projection for Energy Optimization
• Interior Point Solving for LP-based prediction+optimisation
• Learning to Solve Network Flow Problems via Neural Decoding
• Implicit differentiation of Lasso-type models for hyperparameter optimization
• On the Iteration Complexity of Hypergradient Computation
• Gradient Estimation with Stochastic Softmax Tricks
• Nonsmooth Implicit Differentiation for Machine Learning and Optimization
• Sinkformers: Transformers with Doubly Stochastic Attention
• Monotone operator equilibrium networks
• CombOptNet: Fit the Right NP-Hard Problem by Learning Integer Programming Constraints
• Enforcing robust control guarantees within neural network policies
• Learning Convex Optimization Control Policies
• LP-SparseMAP: Differentiable Relaxed Optimization for Sparse Structured Prediction
• Almost Surely Stable Deep Dynamics
• To The Point: Correspondence-driven monocular 3D category reconstruction
• Solving Structured Hierarchical Games Using Differential Backward Induction
• The Differentiable Cross-Entropy Method
• Differentiable Trust Region Layers for Deep Reinforcement Learning
• Extending Lagrangian and Hamiltonian Neural Networks with Differentiable Contact Models
• Automatically Learning Compact Quality-aware Surrogates for Optimization Problems
• On Training and Evaluation of Neural Network Approaches for Model Predictive Control
• Constrained Feedforward Neural Network Training via Reachability Analysis
• Graph Neural Networks for Scalable Radio Resource Management: Architecture Design and Theoretical Analysis
• Deep Random Splines for Point Process Intensity Estimation of Neural Population Data
• From Predictions to Decisions: Using Lookahead Regularization
• Adversarially Robust Learning for Security-Constrained Optimal Power Flow
• Differentiable Equilibrium Computation with Decision Diagrams for Stackelberg Models of Combinatorial Congestion Games
• Escaping from Zero Gradient: Revisiting Action-Constrained Reinforcement Learning via Frank-Wolfe Policy Optimization
• Learning Revenue-Maximizing Auctions With Differentiable Matching
• Distributionally Robust Weighted $k$-Nearest Neighbors
• Differentiable Implicit Layers
• Optimization Induced Equilibrium Networks
• Multiset-Equivariant Set Prediction with Approximate Implicit Differentiation
• End-to-End Balancing for Causal Continuous Treatment-Effect Estimation
• Connecting What to Say With Where to Look by Modeling Human Attention Traces
• Differentiable Learning Under Triage
• Learning the Step-size Policy for the Limited-Memory Broyden-Fletcher-Goldfarb-Shanno Algorithm
• Super-efficiency of automatic differentiation for functions defined as a minimum
• Learning Dynamics Models with Stable Invariant Sets
• Differentiable Forward and Backward Fixed-Point Iteration Layers
• Probabilistic model predictive safety certification for learning-based control
• LCollision: Fast Generation of Collision-Free Human Poses using Learned Non-Penetration Constraints
• Neural Fixed-Point Acceleration for Convex Optimization
• Proximal Mapping for Deep Regularization
• Robust Implicit Networks via Non-Euclidean Contractions
• Portfolio Construction Using Stratified Models
• Strategic Classification Made Practical
• Differentiable Spline Approximations
• Distributional Robust Kelly Gambling: Optimal Strategy under Uncertainty in the Long-Run
• A Surrogate Objective Framework for Prediction+Optimization with Soft Constraints
• Differentiating through the Fréchet Mean
• Deep Likelihood Network for Image Restoration with Multiple Degradation Levels
• End-to-End Learning and Intervention in Games
• Differentiable Robust LQR Layers
• Unfolding Projection-free SDP Relaxation of Binary Graph Classifier via GDPA Linearization
• A Differentiable Recipe for Learning Visual Non-Prehensile Planar Manipulation
• Incremental Proximal Multi-Forecast Model Predictive Control
• Deterministic Certification to Adversarial Attacks via Bernstein Polynomial Approximation
• Learning Models of Model Predictive Controllers using Gradient Data
• Meta-Forecasting by combining Global Deep Representations with Local Adaptation
• Learning-Augmented Sketches for Hessians