Want to create interactive content? It’s easy in Genially!

Get started free

Deep Learning Quiz

Maxime morel

Created on April 7, 2023

Start designing with a free template

Discover more than 1500 professional designs like these:

Meet the Team Game Mobile

Meet the Team Game

Cognitive Stimulation Games

Question - Answer

Character Clue Game

Secret Clue Game

Millionaire Quiz

Explore all templates

Transcript

Turn up the sound !

Are You Full ? Sobriety QUIZ

TRIVIAL

Are you seeing double ?

Are you seeing double ?

START

Top of the hill !

question 1/5 - DEFINITION

What is deep learning ?

Artificial neural network simulated in computer

A drill compagny

Zodiac sign

QUESTION 1/5 - definition

It's a bit cloudy !

You are high, but you don't let go !

Explanation :

  • Machine learning techniques where several "layers" of simple processingunits are connected in a network

question 2/5 - Origins

Don't fall now !

Who try to model networks of neuron with computational circuits ?

John McCarthy

Darth Vader

Mc Culloch&Pitts (1943)

question 2/5 - Origins

On my way to middle earth ...

Good idea, drop the glass !

Explanation :

  • Perceptron algorithm

question 3/5 - Neural Networks

Come onJack !

What is the Weights in neural networks ?

Your balance weight

A data model

A set of adaptive parameters

question 3/5 - Neural networks

It'sfreezinghere !

Your brain is on fire right now !

Explanation :

  • multiple models based on a logistic regression, i.e.linear function + a non-linear function
  • Used as multipliers on the inputs of the neuron, which are added up

question 4/5 - Training and cost

Nearthe end ...

Wich learning algorithm is the fastest to train ?

Gradient

Vanishing gradient

Backpropagation

QUESTION 4/5 - Training and cost

Glass half empty, glass half full !

Heavyhammer ...

Explanation :

  • Rate at which the cost changes with respect to a change in a weight or a bias
  • Gradient at a layer = multiplication of gradient at prior layers

question 5/5 - Convolutional neural networks

Which of this methods is not suitable to decrease the DNN Complexity ?

dryness ...

Parameter quantisation and pruning

Knowledge distillation

Reduce weights parameter

question 5/5 - Convolutional neural networks

Now you're sober, cool nah ?

Explanation :

  • 4 methods :
    • Parameter quantisation and pruning
    • Compressed convolutional filters and matrix factorisation
    • Network architecture search
    • Knowledge distillation

One,two, three ...

Congratulation !

You finished the quiz !

Not the good way, go back !

What a miss !

Do it again !

  • Take a breath ...
  • Be calm ...
  • Grab your water bottle ...
  • You can't make it !