286KiB, 1309x930, ICLH_Diagram_Batch_01_03-DeepNeuralNetwork-WHITEBG.jpg
I just watched the 3Blue1Brown videos on neural networks. https://www.youtube.com/watch?v=aircAruvnKk
He explains neural networks using a character recognition example, and says that the network doesn't actually "learn" what it is that makes the character "8" an 8. If I have understood correctly, machine learning is nothing more than a slightly more complicated curve fitting algorithm. You tune the network coefficients by giving it a ton of inputs, and you hope that in actual use cases, your inputs will be similar enough to the training data for the algorithm to spit out a correct answer. Is this right? If it is, then why the FUCK is there so much excitement amount meme learning?
He explains neural networks using a character recognition example, and says that the network doesn't actually "learn" what it is that makes the character "8" an 8. If I have understood correctly, machine learning is nothing more than a slightly more complicated curve fitting algorithm. You tune the network coefficients by giving it a ton of inputs, and you hope that in actual use cases, your inputs will be similar enough to the training data for the algorithm to spit out a correct answer. Is this right? If it is, then why the FUCK is there so much excitement amount meme learning?
