learningr | Helpful resources for learning R | Frontend Framework library

I want to use some of the parameters of the original deepnet package, so I set up a custom model. I read Caret's documentation (Custom Model), but it doesn't work.

Here is my code for setting up the customized model:

I'm having some trouble following a guide at: https://www.geeksforgeeks.org/ml-neural-network-implementation-in-c-from-scratch/ I have installed the eigen library with vcpkg and it seems to be working because it gives no error.

Code:

I am playing with DL4J version 1.0.0-beta3 and trying to create a convolutional neural network for recognizing 32x32 images of chess pieces. Here is the code I use to create and train the net:

I am trying to write my first neural network to play the game connect four. Im using Java and deeplearning4j. I tried to implement a genetic algorithm, but when i train the network for a while, the outputs of the network jump to NaN and I am unable to tell where I messed up so badly for this to happen.. I will post all 3 classes below, where Game is the game logic and rules, VGFrame the UI and Main all the nn stuff.

I have a pool of 35 neural networks and each iteration i let the best 5 live and breed and randomize the newly created ones a little. To evaluate the networks I let them battle each other and give points to the winner and points for loosing later. Since I penalize putting a stone into a column thats already full I expected the neural networks at least to be able to play the game by the rules after a while but they cant do this. I googled the NaN problem and it seems to be an expoding gradient problem, but from my understanding this shouldn't occur in a genetic algorithm? Any ideas where I could look for the error or whats generally wrong with my implementation?

Main

I am complete beginner to AI as well as tensorflow.js. Currently following the Machine Learning course of Stephen Grider. I should have got a output after the following code but instead i got error. Please help:

code: linear-regression.js:

I am working on an image recognition neural network with Pytorch. My goal is to take pictures of handwritten math equations, process them, and use the neural network to recognize each element. I've reached the point where I am able to separate every variable, number, or symbol from the equation, and everything is ready to be sent through the neural network. I've trained my network to recognize numbers quite well (this part was quite easy), but now I want to expand the scope of the neural network to recognizing letters as well as numbers. I loaded handwritten letters along with the numbers into tensors, shuffled the elements, and put them into batches. No matter how I vary my learning rate, my architecture (hidden layers and the number of neurons per layer), or my batch size I cannot get the neural network to recognize letters.

Here is my network architecture and the feed-forward function (you can see I experimented with the number of hidden layers):

I need a help for my following problem. I'm trying to feed my csv data to my first layer which is convolution1d but it shows

Input 0 is incompatible with layer conv1d_Conv1D1: expected ndim=3, found ndim=2

Here is my code

can someone explain me why I cant change the learningrate during training, in the old Optimizer I could change it with self.updates.append(K.update(self.learning_rate, new_learning_rate)) but cant do it anymore and the self._set_hyper("learning_rate", new_learning_rate) doesn't work, it tells me that: TypeError:__array__() takes 1 positional argument but 2 were given

I made a simple NN predict x from Sin(x). It failed. The NN was successful in predicting sin(x) form x but could not predict x from Sin(x). in both cases(sin(x) and arcsin(x)) we have a non-linear mapping and NN is supposed to be able to fit any function. so, my question is why the NN failed? is this a case of underfitting? can I figure out at which point in the training process the divergence happens?