Here is a very easy to use and powerful Neural Network AI sprite you can use in your projects. If you know about NN's then you will find this project great cause its built the way you think it should unlike many others. You can create as many AI as you want as long as the size of all the AI combined is not larger than the 200,000 item limit in lists. Typically sized AI will look something like this: [10, 10, 5, 2] or [2, 10, 1]. In these cases the total size of the network will be 194 parameters (224 with other values like name and such) and 52 parameters (70). So you could have 892 of the first AI and 2857 of the second AI. Or you can go much bigger and have an AI like this: [64, 64, 64, 64] which would be 12672 (12756) so you could have 15 of those. It starts to get slow at that size so TurboWarp will fix that!
Feature Modules (Organization Of Function): - Create - Shape - Process - Data - Modify - Inputs - Outputs - Tools - Foreach Feature Modules Breakdown: - Create - [[ Create AI Network using the List (AI | Shape). Each item in the list corresponds with the size of that layers nodes ]] - Shape - [[ You should clear the list before adding layer sizes. You can use the functions to clear and add to that shape list easily ]] - Process - [[ The actual calculation of the network. Running this with an AI's index will take the inputs of that AI provided by the Inputs module and run it though the network and creates a prediction/thought of the model. You can use the Outputs module to get those outputs from that AI ]] - Data - [[ Functions for getting data from the AI like the size of the data and name...etc ]] - Modify - [[ Functions to modify the network in some way. Mutate function will change ALL the values of the AI's parameters randomly by the amount given (typically you would use an amount of 0.1 for a little change so that AI is similar to what it was before), the Clamp Parameters function will clamp those same values between the min and max. Also copy the "AI | Modify" function code and replace the "ask" functions to make your own modifications on each parameter in the AI ]] - Inputs - [[ functions to help you access the inputs of the AI given. This is so you can easily set input values in a separate list and push that to the AI's input values ]] - Outputs - [[ functions to help you access the outputs of the AI given. This is so you can easily loop over the outputs values of an AI in a separate list ]] - Tools - [[ these are functions for helping other functions and such. Things like activation function: tanh and leaky relu...etc are here ]] - Foreach - [[ This module shows you efficient AI looping. Copy the code provided in the "AI | Foreach" function. It utilizes data length for optimization, allowing you to loop over the AI list effectively ]]
