Neural networks Flashcards
(21 cards)
What is a neural network?
A set of neurons connected by directed weighted edges
What is a neuron?
A fundamental processing element
What is a dendrite?
A set of inputs
What is an axon
A single output
What is a synapse?
A set of weights
How does a neural network work?
If the sum of inputs >= threshold decided, then the neuron is fired
What does a positive weight mean in a neural network?
Encourages a network to fire
What does a negative weight mean in a neural network?
Prevents a neuron from firing
What is an epoch?
The entire training set fed into the neural network
What is the training value?
The value that we require the network to product
What is the learning rate?
How fast the network converges to an output by adjusting the scale to correct the weights
What happens if the learning rate is too high?
Not able to predict accurately and skips the optimal solution
What happens if the learning rate is too low?
Not able to predict accurately and the training time increases
What do we do when the epoch produces an error?
- Check the next inputs
- Calculate the error by doing training value - output
- If the error is not 0, then weight = weight + learningRateinputserror
What does it mean if a system is linearly separable?
The output data points can be separated using a linear boundary
Explain what the values on a confusion matrix are
- True positive
- True negative
- False negative
- False positive
What does it mean to be true negative?
Model predicts correctly for negative results
What does it mean to be true positive?
Model predicts correctly for positive results
What does it mean to be false negative?
Model predicts incorrectly for negative results
What does it mean to be false positive?
Model predicts incorrectly for positive results
How do we create XOR using other logic?
A XOR B = (A AND NOT B) OR (B AND NOT A)
- We have a hidden neural network layer
