The Mathematics Behind Neural Networks — Concept Check
Questions on The Mathematics Behind Neural Networks. A neural network is a stack of linear transformations and non-linear activations, trained by gradient descent using backpropagation — this concept assembles every earlier topic into one working picture.
Questions in this quiz
8 questions on The Mathematics Behind Neural Networks — Concept Check. Work through them here, then start the quiz to answer them and see your score.
1.A neural network model returns a probability distribution as output. Which function is most likely used in the final layer to achieve this?
- ASigmoid
- BSoftmax
- CReLU
- DTanh
2.During training, a neural network's accuracy plateaus despite a decreasing loss. What could be a reason for this?
- AOverfitting to the training data
- BGradient vanishing problem
- CUnderfitting to the training data
- DLearning rate is too high
3.A neural network uses ReLU activations. What issue might you encounter during training?
- AOverfitting
- BDying ReLU problem
- CVanishing gradients
- DExploding gradients
4.You notice your model's weights are not updating during training. What might be the cause?
- AToo many layers
- BToo few neurons
- CIncorrect loss function
- DLearning rate is too low
5.A network's output layer uses a softmax function. What property does this give the output?
- AOutputs are all positive
- BOutputs are binary
- COutputs sum to 1
- DOutputs are zero-centered
6.In the forward pass of a neural network, what is the role of non-linear activation functions?
- APrevent overfitting
- BEnsure the output is a probability
- CEnable the network to approximate complex functions
- DSpeed up training
7.A neural network's loss decreases, but the accuracy doesn't improve. What might be happening?
- AData is not shuffled
- BLoss function is not suitable
- CModel is overfitting
- DLearning rate is too high
8.What is a common mistake when initializing weights in a neural network?
- AUsing a large batch size
- BUsing a small learning rate
- CUsing random values
- DSetting all weights to zero