+- Small Basic Forum (https://litdev.uk/mybb)
+-- Forum: Small Basic (https://litdev.uk/mybb/forumdisplay.php?fid=1)
+--- Forum: Discussion (https://litdev.uk/mybb/forumdisplay.php?fid=4)
+--- Thread: Creation of SB-Neuron. Ours. Branded.(v2) (/showthread.php?tid=133)
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-08-2024
(10-08-2024, 04:00 PM)jrmrhrb00 Wrote: AB,
I will attach a file for an or gate. I haven't figured out how to do multiple gates. I couldn't do an attachment. So, I put it in as code/
...
Dear JR,
let's go together one step at a time.
Step one.
When you train a neural network to perform operations of only one type (for example, "or"), then the neural network perceives all input data as data FOR THE "OR" OPERATION.
If you want your neural network to work correctly with mixed data of different types, then all input data (both training and working) must additionally contain information ABOUT the TYPE of operation for which these data are intended.
That is, if, while using a neural network, you supply 0 and 1 to the input, then the neural network must also know FOR WHICH OPERATION these input data are intended.
To do this, you will have to add another input to the neural network.
The OPERATION CODE that the neural network must perform with the data that will be at that moment at the other two inputs will be fed to this input.
That is, if the code "0" denotes the "AND" operation, the code "1" denotes the "OR" operation, and the code "2" denotes the "NOT" operation, then the input combination of numbers "1,0,1" will mean "Perform the OR operation on the numbers 0 and 1.".
A neural network trained on such data will be able to work successfully on input data that is organized in the same way.
This is the first step.
RE: Creation of SB-Neuron. Ours. Branded.(v2) - jrmrhrb00 - 10-08-2024
AB,
I think I follow this so far. Add a 3rd input and use it's value to determine the inputs for inputs 1 and 2.
JR
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-09-2024
(translated by Google translator)
Hello.
JR, don't forget that if you use a classic neural network, the operation code must be a normalized number that is in the range from 0 to 1.
For example: 0.1; 0.2; 0.3 and so on.
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-09-2024
To LitDev,
Hi.
If we use a matrix of pixels of size 11 by 11 and we want the SB-Neural Network to learn to recognize "crosses" and "zeros" of size 5 by 5 pixels on this matrix, will this task be called "Image Recognition"?
In addition, we can train the neural network by placing images only on three-quarters of the matrix surface. But, during the use of the neural network, we will place images even in the area of the matrix in which the image was never in the training data.
What will you say?
Is this task suitable for studying the capabilities of our SB-Neural Network?
RE: Creation of SB-Neuron. Ours. Branded.(v2) - litdev - 10-09-2024
AB,
Yes this is a good test, there must be thousands of possible x or o possible and millions of images that are neither.
For me, the first task would be to somehow create a training set and a test test of images that are not the same, there may be some already on the Web somewhere as this is a good example that must have already been attempted. You may even be able to get AI to create them for you.
https://e2eml.school/how_convolutional_neural_networks_work.html
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-09-2024
(10-09-2024, 09:50 AM)litdev Wrote: AB,
...
https://e2eml.school/how_convolutional_neural_networks_work.html
LitDev,
thanks for this link.
Until today, I didn't want to read additional articles about neural networks.
I was interested in exploring the unknown myself.
But today I am already interested in COMPARE my data about the neural network with data from other sources.
I am starting to read.
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-09-2024
LitDev,
I read the article you linked to.
I was a little disappointed that this article is about a neural network. Right now I’m still fascinated by a single neuron. I haven’t yet gotten to the point of consciously purposefully creating a network of individual neurons.
However, the task of recognizing "crosses" and "zeros" remains important to me. Thinking about this problem, I hope to find new thoughts for a separate SB-Neuron.
In short, the fun continues!..
RE: Creation of SB-Neuron. Ours. Branded.(v2) - litdev - 10-09-2024
I guess AI is about the interaction of multiple neurons, not a single neuron (which is just a single function).
RE: Creation of SB-Neuron. Ours. Branded.(v2) - AbsoluteBeginner - 10-09-2024
(10-09-2024, 05:38 PM)litdev Wrote: I guess AI is about the interaction of multiple neurons, not a single neuron (which is just a single function).
I absolutely agree with you.
But I feel uncomfortable if I imagine that I have to create a system from elements whose properties I do not understand.
Therefore, I first began to study the neuron.
RE: Creation of SB-Neuron. Ours. Branded.(v2) - z-s - 10-10-2024
Best website for studying ai.
https://www.w3schools.com/ai/default.asp
The neuron we call are called perceptrons in ai.
https://www.w3schools.com/ai/ai_perceptrons.asp
I also study the whole csharp from this website.
While I am revealing my new project in csharp.
Sharp basic: I am trying to make java script as simple as small basic.
Just using a transpilier because web development is the thing which could be run on any platform.
For that I have to learn js and html first.