I am a high-school students who is learning about data science in his free time. I have gotten a neural network to work which is able to solve xor problems. My neural network uses sigmoid as the activation function for both the hidden and output layers. It also has only one hidden layer. I am wondering about what would be the best simple problem which I could solve with my neural net. I would like a database in which there is a probability output or something similar since I've had problems converting sigmoided output to normal values. I have looked on the UCI machine learning repository but have found nothing witch has caught my eye. I would appreciate any help! :)
Search "dataset" instead of "database".
Try working with MNIST dataset if you are new to neural networks. Also, converting sigmoid function outputs probabilities which you can convert to class labels using argmax function. For ex: If you have 2 classes to classify then your network will output probability for true class using sigmoid which you can convert to labels as 0/1 using argmax. Hope it helps!