Machine learning is often trial and error

Front

To visualize this, imagine that  someone blindfolds you. They tell you that your assignment is to climb a hill and reach its exact top using the fewest number steps possible. Then, they send you walking up the hill. 

You start by measuring off strides that you consider small, medium, and large. Which might be the best way to most closely reach the top of the hill?

Back

Begin with a random size step, then try smaller and larger steps until you find the best stride length

(With very tiny strides, you might reach the top but, it could take days to get there. Broad strides or leaps would get you there quickly, but you’d probably overshoot the top. The best solution might be to begin with a random size step, see where it takes you, then try again with smaller or larger steps until you find the optimum solution.)

Front

After calculating values, what do neural networks often assign to those values that impact the final result?

Back

Weight

(Algorithm results are often assigned a weight that raises or lowers their impact on the final result.)

Front

Fill in the blank. In deep learning, many forms of _______________ devices make up the deep learning ecosystem.

Back

Deep Neural Network (DNN)

(Many forms of deep neural network (DNN) devices make up the modern deep learning ecosystem.)

Front

What kind of brain cells inspired the creation of neural networks?

Back

Neurons

(Neural networks were inspired by the complex way neurons communicate in the human brain.)

Front

You are designing a perceptron. Where will you locate the algorithms that run on the signal and then next passes the results to the output layer?

Back

In the hidden layers

(You would place them in the perceptron’s hidden layers. A perceptron has an input layer, one or more hidden layers, and an output layer. A signal enters the input layer and the hidden layers run algorithms on the signal. Then, the result is passed to the output layer.)

Front

How can you demonstrate the concept of machine learning?

Back

Determine the farthest distance from a target that an archer can reliably hit bullseyes by shooting arrows repeatedly while walking closer and closer to the target.

(Shooting arrows at a target from gradually shorter distances could describe machine learning’s way of performing a series of calculations (similar to shooting arrow at different distances) and noting correlated results in order to increase the accuracy of its algorithms.)

Front

In which model does the generator try to create data that's realistic enough to fool the discriminator, but the discriminator learns to distinguish between real and generated data?

Back

Generative adversarial network (GAN)

(This is how the GAN model works. This competition between the generator and discriminator leads to the generator creating increasingly realistic content.)

Front

True or false? Roberto works as a coach on a professional sports team. One of his players is returning to workouts with the team after being out 6 months due to an injury. Roberto could use generative AI to create a customized workout plan tailored to the returning athlete's current fitness level.

Back

TRUE

(By analyzing an athlete’s biometric data, generative AI can create customized workout plans that are tailored to the athlete’s fitness level and goals. Therefore, Roberto could use generative AI as a tool to help create a customized workout plan for the returning athlete.)

Front

Example of generative AI

Back

Composing music

(Generative AI is a type of artificial intelligence that creates new, original content that has never been seen before. Therefore, composing music is an example of creating new, original content.)

Front

A distinguishing feature of generative AI, as opposed to other types of AI, is its ability to _______________.

Back

Create new content

(Generative AI is a type of artificial intelligence that creates new, original content that has never been seen before.)