Perceptron returning values larger than 1 and smaller than 0 + sigmoid func

Help and Feedback Scripting Support
1

Hello, recently I have started to look into Neural netwrosk and I have finally decided to create my first one. It is meant to predict your/my gender based on height and weight, unfortunelyy i could not find an api with large data on female and males weight and height to I have just created a view table containing
their weight in kg, their height in cm and gender 1 as male and 0 as female.

Thats some context the real issue is I am passing the sum of the two imputs and weights plus bias into a sigmoid function, but it is still returing values that are larger than 1 and smaller than zero.

Here is my sigmoid function:

local S = function(x)
	return 1 / (1 + math.exp(-x))
end

And this is the rest of my code:

local S = function(x)
	return 1 / (1 + math.exp(-x))
end

local Randomnew = Random.new()

local w1 = Randomnew:NextNumber(0, 1)
local w2 = Randomnew:NextNumber(0, 1)

local b = Randomnew:NextNumber(0, 1)

local r = .1

local Perceptron = function(Inputs, Output)
	local h1 = S(w1 * Inputs[1] + w2 * Inputs[2] + b)
	warn(h1)
	if Output then
		local loss = Output - h1

		w1 += loss * Inputs[1] * r
		w2 += loss * Inputs[2] * r
	end
	return h1
end

local Averages = {
	--Males:
	{
		184,
		87.9,
		1
	},
	{
		183,
		90.4,
		1
	},
	{
		182,
		89.9,
		1
	},
	{
		182,
		86.8,
		1
	},
	{
		182,
		87.1,
		1
	},
	--Females:
	{
		170,
		73.2,
		0
	},
	{
		170,
		75.3,
		0
	},
	{
		168,
		73.7,
		0
	},
	{
		169,
		70.2,
		0
	},
	{
		167,
		70.6,
		0
	}
}

local MyGender = Perceptron({
	176,
	60.2
})
print("Before:", MyGender)

for Index = 1, 100 do
	table.foreachi(Averages, function(_, Value)
		Perceptron(Value, Value[3])
	end)
end

local MyGender = Perceptron({
	176,
	80.2
})
print("After:", MyGender)

Basically I do not understand why I am getting large and negative values eveen though they have been based through my sigmoid function

Any help at all is appreciated thanks.

2

Solved it just need to divide male and females weights and heights by their averages.

3

This csv has enough data that it appears to be a normal distribution.
image

Just parse this CSV file.

local rows = data:split("\n")
table.remove(rows, 1)
local training = {}
for i, row in pairs(rows) do
    local gender, heightInches, weightPounds = table.unpack(row:split(","))
    table.insert(training, { gender == [["Male"]] and 1 or 0, heightInches * 2.54, weightPounds / 2.205 })
end
4

Thanks for the info, when I have become confident with feedForward networks I will try and create a self driving car