ulfaslak / annadl_f19 Goto Github PK

Don't we want one prediction value after calling the feedforward function on a single point of our x vector? I am getting an array of 3 values. Does anyone know how to somehow squish this into 1 or are we supposed to work with 3?

I'm having an issue where after finally finding parameters for the NN with which it overfits the random data (noise), if I regenerate the data the NN no longer trains to overfit (gets stuck around 50% accuracy). Then I have to regenerate the data many times to end up with noise that the NN can overfit again. Anyone with the same issue?

(Using 1000 neuron hidden layer, learning rate of 0.05, momentum of 0.9, relu activation, 50 batch size, 10,000 epochs)

Can anybody say what the difference is between the two vectors in update_mini_batch?

net1 = Network([2,1])

X, y = generate_X_linear(1e5)
X, y = prepare_data(X, y)

gives me this error: "ValueError: too many values to unpack (expected 2)."

When I pass it X[:2], y[:2] it spits out the respective tuple, but it can't digest the entire dataset. Any ideas?

I'm getting the following error when running the nn.Network(512, 2):

KeyError: 2

I've tried googling it, but can't find a meaningful result. Anyone knows what's going on here?

My code reads:

from torch import nn
import numpy
nn.Network(512, 2)