New loss #937

AstitvaAggarwal · 2025-04-13T07:16:59Z

Checklist

Appropriate tests were added
Any code changes were done in a way that does not break public API
All documentation related to code changes were updated
The new code follows the
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Any new documentation only uses public API

Additional context

Add any other context about the problem here.

AstitvaAggarwal · 2025-04-27T20:53:25Z

@ChrisRackauckas i think there are some compat issues in the Downgrade Test env.

ChrisRackauckas · 2025-04-27T21:10:39Z

Yes don't worry about downgrade

AstitvaAggarwal · 2025-05-05T18:47:23Z

@ChrisRackauckas the following PR adds the new loss for NNODE and BNNODE (with appropriate tests) and corrects tests erroring out in BPINN_PDE_tests.jl (this started once the repo was completely overhauled). Just to give some insight, BPINN model performance for just 20 training points in t=(0,4) as in the tests added where we solve LV :

u2 is our new model.

src/BPINN_ode.jl

ChrisRackauckas · 2025-05-06T14:56:43Z

src/ode_solve.jl

+        # Quadrature is applied on timewise losses
+        # Gridtraining/trapezoidal rule quadrature_weights is dt.*ones(T, length(t))
+        return sum(sum(abs2, loss_vals[i, :] .* quadrature_weights) for i in 1:n_output)


How is this different from the quadrature loss?

is there a reason to not cubature?

its pretty much a quadrature loss but we dont create an Integralproblem as we have a fixed set of precomputed points to get a loss function. Not using HCubature here as again we cant compute f(x) at arbitrary domain points (so its not exactly h or p adaptive integration)

But why not combine the two loss calculations and cubature it? That would be strictly faster convergence?

ohh okay, missed that, thanks !

Co-authored-by: Christopher Rackauckas <[email protected]>

…E.jl into sdepinn

AstitvaAggarwal · 2025-05-09T14:45:22Z

@ChrisRackauckas GTM?

src/ode_solve.jl

ChrisRackauckas · 2025-05-12T10:34:51Z

src/advancedHMC_MCMC.jl

@@ -6,7 +6,7 @@
    dataset <: Union{Vector{Nothing}, Vector{<:Vector{<:AbstractFloat}}}
    priors <: Vector{<:Distribution}
    phystd::Vector{Float64}
-    phynewstd::Vector{Float64}
+    phynewstd::Function


Specialize?

ChrisRackauckas · 2025-05-12T10:52:41Z

src/ode_solve.jl

@@ -91,14 +91,17 @@ Networks 9, no. 5 (1998): 987-1000.
    strategy <: Union{Nothing, AbstractTrainingStrategy}
    param_estim
    additional_loss <: Union{Nothing, Function}
+    dataset <: Union{Vector{Nothing}, Vector{<:Vector{<:AbstractFloat}}}


Use an empty vector as no dataset to get type stable.

ChrisRackauckas · 2025-05-12T10:52:53Z

src/ode_solve.jl

@@ -91,14 +91,17 @@ Networks 9, no. 5 (1998): 987-1000.
    strategy <: Union{Nothing, AbstractTrainingStrategy}
    param_estim
    additional_loss <: Union{Nothing, Function}
+    dataset <: Union{Vector{Nothing}, Vector{<:Vector{<:AbstractFloat}}}


Need to document in the docstring the required dataset form and the new kwargs

ChrisRackauckas · 2025-05-12T10:54:48Z

test/NNODE_tests.jl

+
+@testitem "ODE Parameter Estimation Improvement" tags=[:nnode] begin
+    using OrdinaryDiffEq, Random, Lux, OptimizationOptimJL, LineSearches
+    using FastGaussQuadrature, PolyChaos, Integrals


Suggested change

using FastGaussQuadrature, PolyChaos, Integrals

using FastGaussQuadrature

ChrisRackauckas · 2025-05-12T10:54:59Z

Project.toml

@@ -92,6 +93,7 @@ Optimization = "4"
 OptimizationOptimJL = "0.4"
 OptimizationOptimisers = "0.3"
 OrdinaryDiffEq = "6.87"
+PolyChaos = "0.2.11"


Suggested change

PolyChaos = "0.2.11"

ChrisRackauckas · 2025-05-12T10:55:05Z

Project.toml

@@ -126,10 +129,11 @@ LuxLib = "82251201-b29d-42c6-8e01-566dec8acb11"
 MethodOfLines = "94925ecb-adb7-4558-8ed8-f975c56a0bf4"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
+PolyChaos = "8d666b04-775d-5f6e-b778-5ac7c70f65a3"


Suggested change

PolyChaos = "8d666b04-775d-5f6e-b778-5ac7c70f65a3"

ChrisRackauckas · 2025-05-12T10:55:14Z

Project.toml

 ReTestItems = "817f1d60-ba6b-4fd5-9520-3cf149f6a823"
 StochasticDiffEq = "789caeaf-c7a9-5a7d-9973-96adeb23e2a0"
 TensorBoardLogger = "899adc3e-224a-11e9-021f-63837185c80f"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

 [targets]
-test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]
+test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "FastGaussQuadrature", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "PolyChaos", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]


Suggested change

test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "FastGaussQuadrature", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "PolyChaos", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]

test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "FastGaussQuadrature", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]

AstitvaAggarwal added 9 commits April 13, 2025 12:40

Too Easy.

29e123e

remove Integrals from test deps

6bf8c49

tests. (new is better in more noise)

45c1e39

minor change

83a642d

.

5cfc6e7

.

d8f602b

Likelihood probabilites are not driven to 0.

30fa615

.

85b350c

more samples

3c59fde

AstitvaAggarwal changed the title ~~Complete Constrained PINNS, BPINNs.~~ New loss Apr 14, 2025

AstitvaAggarwal and others added 10 commits April 14, 2025 20:54

.

f2aafdb

fixed tests

70e956d

tests.

41691eb

.

f21969c

std for new loss is parametric

a968cc8

Changes to API

c66772a

tests.

2fcf75e

tests-2

cb85411

tests-3

2423224

Update BPINN_tests.jl

cc1348f

AstitvaAggarwal added 6 commits April 28, 2025 03:10

BPINN_PDE loss corrected

4a1ca0e

NNODE improvements & L2Data!=additional_loss

107165e

spelling check

a4d1fb7

tests

d47c19c

tests-1

0bc0ec1

tests-3

65c4b08

format

8ed4e18

ChrisRackauckas reviewed May 6, 2025

View reviewed changes

src/BPINN_ode.jl Outdated Show resolved Hide resolved

ChrisRackauckas reviewed May 6, 2025

View reviewed changes

AstitvaAggarwal and others added 4 commits May 6, 2025 21:16

Update src/BPINN_ode.jl

c242419

Co-authored-by: Christopher Rackauckas <[email protected]>

cubature over L2 instead of L1

95140dd

Merge branch 'sdepinn' of https://github.com/AstitvaAggarwal/NeuralPD…

6cb24c5

…E.jl into sdepinn

bpinn remains in non squared within logpdf?

f24df29

ChrisRackauckas reviewed May 12, 2025

View reviewed changes

src/ode_solve.jl Show resolved Hide resolved

ChrisRackauckas reviewed May 12, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

New loss #937

New loss #937

AstitvaAggarwal commented Apr 13, 2025

AstitvaAggarwal commented Apr 27, 2025

ChrisRackauckas commented Apr 27, 2025

AstitvaAggarwal commented May 5, 2025 •

edited

Loading

ChrisRackauckas May 6, 2025

ChrisRackauckas May 6, 2025

AstitvaAggarwal May 6, 2025 •

edited

Loading

ChrisRackauckas May 6, 2025

AstitvaAggarwal May 6, 2025 •

edited

Loading

AstitvaAggarwal commented May 9, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

ChrisRackauckas May 12, 2025

	using FastGaussQuadrature, PolyChaos, Integrals
	using FastGaussQuadrature

	test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "FastGaussQuadrature", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "PolyChaos", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]
	test = ["Aqua", "CUDA", "DiffEqNoiseProcess", "ExplicitImports", "FastGaussQuadrature", "Flux", "Hwloc", "InteractiveUtils", "LineSearches", "LuxCUDA", "LuxCore", "LuxLib", "MethodOfLines", "OptimizationOptimJL", "OrdinaryDiffEq", "ReTestItems", "StochasticDiffEq", "TensorBoardLogger", "Test"]

New loss #937

Are you sure you want to change the base?

New loss #937

Conversation

AstitvaAggarwal commented Apr 13, 2025

Checklist

Additional context

AstitvaAggarwal commented Apr 27, 2025

ChrisRackauckas commented Apr 27, 2025

AstitvaAggarwal commented May 5, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AstitvaAggarwal May 6, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AstitvaAggarwal May 6, 2025 • edited Loading

Choose a reason for hiding this comment

AstitvaAggarwal commented May 9, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AstitvaAggarwal commented May 5, 2025 •

edited

Loading

AstitvaAggarwal May 6, 2025 •

edited

Loading

AstitvaAggarwal May 6, 2025 •

edited

Loading