11.2 TensorFlow

Directions: A Neural Network is one example of a "learning algorithm” that we have been referencing. They were built to loosely resemble the brain and the neurons that pass information through it. 

As our brains learn they strengthen neurons that get to the right answers or get rid of neurons that don’t create good pathways. Our brains are always “learning” and growing. 

Neural networks train on labeled data (supervised learning!) in order to build a model that can make the computer more intelligent. As an AI engineer you are choosing the shape and structure of the neural network. What do we mean when we say shape and structure? We mean how many inputs and outputs, or how many layers or nodes.  We are going to try it ourselves by playing with this tensorflow playground! Follow the instructions below.

Submission: When finished, submit your screenshot in Moodle.

Questions and Information:

• Go here: https://bit.ly/playgroundnn 

• Start with one input and one neuron in the hidden layer. 

• Choose linear regression as your problem type 
• Choose the line as your data set
• Push play and see what happens. (Hint - it should fit the data pretty fast!)
• Now choose a different data set but don’t change other variables and push play 
  
  • Can the network do it?
  • What would you need to change?
  • Try adding neurons or layers in order to make your new model fit!
• Screenshot the new model fitting your chosen data and submit it within Moodle.