DL Flashcards
Explain the difference between Artificial Intelligence, Machine Learning and Deep Learning.
- AI is a technique that enables a machine to mimic human behavior.
- Machine Learning is a technique to achieve AI through algorithms trained with data.
- Deep Learning is a type of Machine Learning inspired by the structure of the human brain, an Artificial Neural Network.
What is the difference when creating a machine to differentiate between tomatoes and cheeries using Machine Learning versus Deep Learning?
ML: de eigenschappen worden meegegeven en op basis daarvan leert de machine het verschil
DL: De machine leert het verschil zonder eigenschappen van bijde mee te krijgen
What are the three limitations or problems when using Deep Learning?
a) Veel data nodig
b) Veel gebruik van CPU
c) Berekeningen nemen veel tijd in beslag
The example neural network that can learn to recognize handwritten digits consists of four layers. What are those four layers used for?
- The first layer consist of 28x28=784 neurons. The value of each neuron is the grey scale of the pixel (ex 0.58).
- The last layer consist of 10 neurons. The value of each neuron corresponds to how much the system thinks that a given image corresponds to the given digit.
- The first hidden layer might try to recognize the various little edges.
- The second hidden layer might try to recognize the patterns the digits are made up from.
What are the weights in a neural network used for? What is a bias used for? Why do we use the sigmoid function?
weight connect our neuron with the neurons of the first layer.
bias tells you how high the weighted sum needs to be before the neuron is activated.
sigmoid is to convert the weighted sum to a value between 0 and 1.
Explain briefly how neural networks are trained.
adjusting the weights and biases of the network to improve the accuracy of the result
Explain the following terms: Gradient Descent and Backpropagation.
- Gradient descent: finding a set of weights and biases which make the cost as small as possible.
- Backpropagation: method for calculating the weights updates to improve the network until it is able to perform the task for which it is being trained.
Declare color marker tracking
The algorith searches pixel per pixel. The pixel with the closest rgb valua to the searched rgb value contains the object.
Declare patches
Small regions of pixels
Declare kernel
contains the values for a pixelwhise multiplication
Computer Vision is one of the areas in AI where core concepts are already being integrated into major products that we use every day. Name four examples.
recognising hand written text
# spotting tumors in ct scans
# monitoring traffic flow on road
# recognising faces
What is a convolution?
a convolution is a mathematical operation between an input image and a filter to produce an outcome or feature map. With this feature map, we detect a particular feature from the input image.
Explain the concept of cascades used in face detection.
cascade breaks the problem of detecting faces into multiple stages. For each block, it does a very rough and quick test. If that passes, it does a slightly more detailed test, and so on.
What are the convolutional layers in a CNN used for?
a convolutional layer receives input, transforms the input in some way, and then outputs the transformed input to the next layer.
Explain the terms padding and stride.
padding means giving additional pixels at the boundary of the data.
Stride is the number of steps a filter moves at a time