🚀 Add NonLinearProgram Support to DiffOpt.jl

Project.toml

@@ -1,7 +1,7 @@
 name = "DiffOpt"
 uuid = "930fe3bc-9c6b-11ea-2d94-6184641e85e7"
-authors = ["Akshay Sharma", "Mathieu Besançon", "Joaquim Dias Garcia", "Benoît Legat", "Oscar Dowson"]
-version = "0.4.3"
+authors = ["Akshay Sharma", "Mathieu Besançon", "Joaquim Dias Garcia", "Benoît Legat", "Oscar Dowson", "Andrew Rosemberg"]
+version = "0.5.0"
 
 [deps]
 BlockDiagonals = "0a1fb500-61f7-11e9-3c65-f5ef3456f9f0"

docs/src/index.md

@@ -1,10 +1,10 @@
 # DiffOpt.jl
 
-[DiffOpt.jl](https://github.com/jump-dev/DiffOpt.jl) is a package for differentiating convex optimization program ([JuMP.jl](https://github.com/jump-dev/JuMP.jl) or [MathOptInterface.jl](https://github.com/jump-dev/MathOptInterface.jl) models) with respect to program parameters. Note that this package does not contain any solver.
+[DiffOpt.jl](https://github.com/jump-dev/DiffOpt.jl) is a package for differentiating convex and non-convex optimization program ([JuMP.jl](https://github.com/jump-dev/JuMP.jl) or [MathOptInterface.jl](https://github.com/jump-dev/MathOptInterface.jl) models) with respect to program parameters. Note that this package does not contain any solver.
 This package has two major backends, available via the `reverse_differentiate!` and `forward_differentiate!` methods, to differentiate models (quadratic or conic) with optimal solutions.
 
 !!! note
-    Currently supports *linear programs* (LP), *convex quadratic programs* (QP) and *convex conic programs* (SDP, SOCP, exponential cone constraints only). 
+    Currently supports *linear programs* (LP), *convex quadratic programs* (QP), *convex conic programs* (SDP, SOCP, exponential cone constraints only), and *general nonlinear programs* (NLP).
 
 
 ## Installation
@@ -16,8 +16,8 @@ DiffOpt can be installed through the Julia package manager:
 
 ## Why are Differentiable optimization problems important?
 
-Differentiable optimization is a promising field of convex optimization and has many potential applications in game theory, control theory and machine learning (specifically deep learning - refer [this video](https://www.youtube.com/watch?v=NrcaNnEXkT8) for more).
-Recent work has shown how to differentiate specific subclasses of convex optimization problems. But several applications remain unexplored (refer section 8 of this [really good thesis](https://github.com/bamos/thesis)). With the help of automatic differentiation, differentiable optimization can have a significant impact on creating end-to-end differentiable systems to model neural networks, stochastic processes, or a game.
+Differentiable optimization is a promising field of constrained optimization and has many potential applications in game theory, control theory and machine learning (specifically deep learning - refer [this video](https://www.youtube.com/watch?v=NrcaNnEXkT8) for more).
+Recent work has shown how to differentiate specific subclasses of constrained optimization problems. But several applications remain unexplored (refer section 8 of this [really good thesis](https://github.com/bamos/thesis)). With the help of automatic differentiation, differentiable optimization can have a significant impact on creating end-to-end differentiable systems to model neural networks, stochastic processes, or a game.
 
 
 ## Contributing

docs/src/manual.md

@@ -1,10 +1,6 @@
 # Manual
 
-!!! note
-    As of now, this package only works for optimization models that can be written either in convex conic form or convex quadratic form.
-
-
-## Supported objectives & constraints - `QuadraticProgram` backend
+## Supported objectives & constraints - scheme 1
 
 For `QuadraticProgram` backend, the package supports following `Function-in-Set` constraints: 
 
@@ -16,6 +12,12 @@ For `QuadraticProgram` backend, the package supports following `Function-in-Set`
 |    `ScalarAffineFunction`    |    `GreaterThan`    |
 |    `ScalarAffineFunction`    |    `LessThan`    |
 |    `ScalarAffineFunction`    |    `EqualTo`    |
+|    `ScalarQuadraticFunction`    |    `GreaterThan`    |
+|    `ScalarQuadraticFunction`    |    `LessThan`    |
+|    `ScalarQuadraticFunction`    |    `EqualTo`    |
+|    `ScalarNonlinearFunction`    |    `GreaterThan`    |
+|    `ScalarNonlinearFunction`    |    `LessThan`    |
+|    `ScalarNonlinearFunction`    |    `EqualTo`    |
 
 and the following objective types: 
 
@@ -24,6 +26,7 @@ and the following objective types:
 |   `VariableIndex`   |
 |   `ScalarAffineFunction`   |
 | `ScalarQuadraticFunction`  | 
+| `ScalarNonlinearFunction`  |
 
 
 ## Supported objectives & constraints - `ConicProgram` backend
@@ -52,7 +55,6 @@ and the following objective types:
 
 Other conic sets such as `RotatedSecondOrderCone` and `PositiveSemidefiniteConeSquare` are supported through bridges.
 
-
 ## Creating a differentiable MOI optimizer
 
 You can create a differentiable optimizer over an existing MOI solver by using the `diff_optimizer` utility. 
@@ -68,7 +70,7 @@ DiffOpt requires taking projections and finding projection gradients of vectors
 ## Conic problem formulation
 
 !!! note
-    As of now, the package is using `SCS` geometric form for affine expressions in cones.
+    As of now, when defining a conic or convex quadratic problem, the package is using `SCS` geometric form for affine expressions in cones.
 
 Consider a convex conic optimization problem in its primal (P) and dual (D) forms:
 ```math

docs/src/reference.md

@@ -4,5 +4,5 @@
 ```
 
 ```@autodocs
-Modules = [DiffOpt, DiffOpt.QuadraticProgram, DiffOpt.ConicProgram]
+Modules = [DiffOpt, DiffOpt.QuadraticProgram, DiffOpt.ConicProgram, DiffOpt.NonLinearProgram]
 ```

docs/src/usage.md

@@ -56,3 +56,65 @@ MOI.set(model, DiffOpt.ForwardObjectiveFunction(), ones(2) ⋅ x)
 DiffOpt.forward_differentiate!(model)
 grad_x = MOI.get.(model, DiffOpt.ForwardVariablePrimal(), x)
 ```
+
+3. To differentiate a general nonlinear program, have to use the API for Parameterized JuMP models. For example, consider the following nonlinear program:
+
+```julia
+using JuMP, DiffOpt, HiGHS
+
+model = Model(() -> DiffOpt.diff_optimizer(Ipopt.Optimizer))
+set_silent(model)
+
+p_val = 4.0
+pc_val = 2.0
+@variable(model, x)
+@variable(model, p in Parameter(p_val))
+@variable(model, pc in Parameter(pc_val))
+@constraint(model, cons, pc * x >= 3 * p)
+@objective(model, Min, 2x)
+optimize!(model)
+@show value(x) == 3 * p_val / pc_val
+
+# the function is
+# x(p, pc) = 3p / pc
+# hence,
+# dx/dp = 3 / pc
+# dx/dpc = -3p / pc^2
+
+# First, try forward mode AD
+
+# differentiate w.r.t. p
+direction_p = 3.0
+MOI.set(model, DiffOpt.ForwardConstraintSet(), ParameterRef(p), Parameter(direction_p))
+DiffOpt.forward_differentiate!(model)
+@show MOI.get(model, DiffOpt.ForwardVariablePrimal(), x) == direction_p * 3 / pc_val
+
+# update p and pc
+p_val = 2.0
+pc_val = 6.0
+set_parameter_value(p, p_val)
+set_parameter_value(pc, pc_val)
+# re-optimize
+optimize!(model)
+# check solution
+@show value(x) ≈ 3 * p_val / pc_val
+
+# stop differentiating with respect to p
+DiffOpt.empty_input_sensitivities!(model)
+# differentiate w.r.t. pc
+direction_pc = 10.0
+MOI.set(model, DiffOpt.ForwardConstraintSet(), ParameterRef(pc), Parameter(direction_pc))
+DiffOpt.forward_differentiate!(model)
+@show abs(MOI.get(model, DiffOpt.ForwardVariablePrimal(), x) -
+    -direction_pc * 3 * p_val / pc_val^2) < 1e-5
+
+# always a good practice to clear previously set sensitivities
+DiffOpt.empty_input_sensitivities!(model)
+# Now, reverse model AD
+direction_x = 10.0
+MOI.set(model, DiffOpt.ReverseVariablePrimal(), x, direction_x)
+DiffOpt.reverse_differentiate!(model)
+@show MOI.get(model, DiffOpt.ReverseConstraintSet(), ParameterRef(p)) == MOI.Parameter(direction_x * 3 / pc_val)
+@show abs(MOI.get(model, DiffOpt.ReverseConstraintSet(), ParameterRef(pc)).value -
+    -direction_x * 3 * p_val / pc_val^2) < 1e-5
+```

src/DiffOpt.jl

@@ -27,6 +27,7 @@ include("bridges.jl")
 
 include("QuadraticProgram/QuadraticProgram.jl")
 include("ConicProgram/ConicProgram.jl")
+include("NonLinearProgram/NonLinearProgram.jl")
 
 """
     add_all_model_constructors(model)
@@ -37,6 +38,7 @@ Add all constructors of [`AbstractModel`](@ref) defined in this package to
 function add_all_model_constructors(model)
     add_model_constructor(model, QuadraticProgram.Model)
     add_model_constructor(model, ConicProgram.Model)
+    add_model_constructor(model, NonLinearProgram.Model)
     return
 end