Skip to content

Test: generalize network for inference unit tests #292

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
rouson merged 2 commits into from
Mar 4, 2026
Merged

Test: generalize network for inference unit tests #292

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
af1b598
feat(activation): add tanh
rouson Mar 4, 2026
c7057ca
test(inference): reduce use of step activation
rouson Mar 4, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions src/fiats/activation_m.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,13 +8,13 @@ module activation_m

private
public :: activation_t
public :: gelu, relu, sigmoid, step, swish
public :: gelu, relu, sigmoid, step, swish, tanh_

enum, bind(C)
enumerator :: gelu=1, relu, sigmoid, step, swish
enumerator :: gelu=1, relu, sigmoid, step, swish, tanh_
end enum

character(len=*), parameter :: activation_name(*) = [character(len("sigmoid")) :: "gelu", "relu", "sigmoid", "step", "swish"]
character(len=*), parameter :: activation_name(*) = [character(len("sigmoid")) :: "gelu", "relu", "sigmoid", "step", "swish", "tanh"]

type activation_t
private
Expand Down
10 changes: 10 additions & 0 deletions src/fiats/activation_s.F90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,8 @@
activation%selection_ = step
case("swish")
activation%selection_ = swish
case("tanh")
activation%selection_ = tanh_
case default
error stop "activation_s(construct_from_name): unknown name"
end select
Expand All @@ -56,6 +58,8 @@
associate(sigmoid_activation => 1./(1.+exp(-x)))
y = x*sigmoid_activation
end associate
case(tanh_)
y = (exp(x) - exp(-x))/(exp(x) + exp(-x))
case default
error stop "activation_s(default_real_evaluate): unknown activation"
end select
Expand All @@ -75,6 +79,8 @@
associate(sigmoid_activation => 1D0/(1D0+exp(-x)))
y = x*sigmoid_activation
end associate
case(tanh_)
y = (exp(x) - exp(-x))/(exp(x) + exp(-x))
case default
error stop "activation_s(double_precision_evaluate): unknown activation"
end select
Expand All @@ -94,6 +100,8 @@
associate(sigmoid_activation => 1./(1.+exp(-x)), sigmoid_differentiate => exp(-x)/(1.+exp(-x))**2)
dy_dx = sigmoid_activation + x * sigmoid_differentiate
end associate
case(tanh_)
dy_dx = 1. - ((exp(x) - exp(-x))/(exp(x) + exp(-x)))**2
case default
error stop "activation_s(default_real_differentiate): unknown activation"
end select
Expand All @@ -113,6 +121,8 @@
associate(sigmoid_activation => 1D0/(1D0+exp(-x)), sigmoid_differentiate => exp(-x)/(1D0+exp(-x))**2)
dy_dx = sigmoid_activation + x * sigmoid_differentiate
end associate
case(tanh_)
dy_dx = 1D0 - ((exp(x) - exp(-x))/(exp(x) + exp(-x)))**2
case default
error stop "activation_s(double_precision_differentiate): unknown activation"
end select
Expand Down
44 changes: 35 additions & 9 deletions test/neural_network_test_m.F90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,24 +43,28 @@ function results() result(test_results)
type(neural_network_test_t) neural_network_test

test_results = neural_network_test%run([ &
test_description_t("performing elemental inference with 1 hidden layer", elemental_infer_with_1_hidden_layer_xor_net) &
,test_description_t("performing elemental inference with 2 hidden layers", elemental_infer_with_2_hidden_layer_xor_net) &
test_description_t("elemental inference with step activation and 1 hidden layer", elemental_infer_1_hidden_layer_xor_step) &
,test_description_t("elemental inference with relu activation and 1 hidden layer", elemental_infer_1_hidden_layer_xor_relu) &
,test_description_t("elemental inference with 2 hidden layers", elemental_infer_with_2_hidden_layer_xor_net) &
,test_description_t("converting a network with 2 hidden layers to and from JSON format", multi_hidden_layer_net_to_from_json) &
,test_description_t("converting a network with varying-width hidden layers to/from JSON", varying_width_net_to_from_json) &
,test_description_t("performing inference with a network with hidden layers of varying width", infer_with_varying_width_net) &
,test_description_t("double-precision inference", double_precision_inference) &
])
end function

function single_hidden_layer_xor_network() result(neural_network)
function one_hidden_layer_xor_net(activation) result(neural_network)
character(len=*), intent(in) :: activation
type(neural_network_t) neural_network
integer, parameter :: nodes_per_layer(*) = [2, 3, 1]
integer, parameter :: max_n = maxval(nodes_per_layer), layers = size(nodes_per_layer)

if (.not. any(activation==["step","relu"])) error stop "neural_network_test_m(one_hidden_layer_xor_net): step or relu required."

neural_network = neural_network_t( &
metadata = [string_t("XOR"), string_t("Damian Rouson"), string_t("2023-07-02"), string_t("step"), string_t("false")], &
metadata = [string_t("XOR"), string_t("Damian Rouson"), string_t("2023-07-02"), string_t(activation), string_t("false")], &
weights = reshape([real:: [1,1,0, 0,1,1, 0,0,0], [1,0,0, -2,0,0, 1,0,0]], [max_n, max_n, layers-1]), &
biases = reshape([[0.,-1.99,0.], [0., 0., 0.]], [max_n, layers-1]), &
biases = reshape([[0.,-1.,0.], [0., 0., 0.]], [max_n, layers-1]), &
nodes = nodes_per_layer &
)
end function
Expand All @@ -71,10 +75,10 @@ function multi_layer_xor_network() result(neural_network)
integer, parameter :: max_n = maxval(nodes_per_layer), layers = size(nodes_per_layer)

neural_network = neural_network_t( &
metadata = [string_t("XOR"), string_t("Damian Rouson"), string_t("2023-07-02"), string_t("step"), string_t("false")], &
metadata = [string_t("XOR"), string_t("Damian Rouson"), string_t("2023-07-02"), string_t("relu"), string_t("false")], &
weights = reshape([real:: [1,1,0, 0,1,1, 1,0,0, 1,0,0, 0,1,0, 0,0,1], [1,0,0, -2,0,0, 1,0,0]], &
[max_n, max_n, layers-1]), &
biases = reshape([[0.,-1.99,0.], [0., 0., 0.], [0., 0., 0.]], [max_n, layers-1]), &
biases = reshape([[0.,-1.,0.], [0., 0., 0.], [0., 0., 0.]], [max_n, layers-1]), &
nodes = nodes_per_layer &
)
end function
Expand Down Expand Up @@ -205,11 +209,33 @@ function double_precision_inference() result(test_diagnosis)
test_diagnosis = .all. (inputs%values() .approximates. outputs%values() .within. tolerance)
end function

function elemental_infer_with_1_hidden_layer_xor_net() result(test_diagnosis)
function elemental_infer_1_hidden_layer_xor_step() result(test_diagnosis)
type(test_diagnosis_t) test_diagnosis
type(neural_network_t) neural_network

neural_network = one_hidden_layer_xor_net("step")

block
type(tensor_t), allocatable :: truth_table(:)
real, parameter :: tolerance = 1.E-08, false = 0., true = 1.
integer i

associate(array_of_inputs => [tensor_t([true,true]), tensor_t([true,false]), tensor_t([false,true]), tensor_t([false,false])])
truth_table = neural_network%infer(array_of_inputs)
end associate
test_diagnosis = &
(.all. (truth_table(1)%values() .approximates. (false) .within. tolerance)) &
.also. (.all. (truth_table(2)%values() .approximates. ( true) .within. tolerance)) &
.also. (.all. (truth_table(3)%values() .approximates. ( true) .within. tolerance)) &
.also. (.all. (truth_table(4)%values() .approximates. (false) .within. tolerance))
end block
end function

function elemental_infer_1_hidden_layer_xor_relu() result(test_diagnosis)
type(test_diagnosis_t) test_diagnosis
type(neural_network_t) neural_network

neural_network = single_hidden_layer_xor_network()
neural_network = one_hidden_layer_xor_net("relu")

block
type(tensor_t), allocatable :: truth_table(:)
Expand Down