URGENT PLEASE HELP. PLEASE MAKE SURE TO INCLUDE SCREENSHOTS WITH…

Question
Answered step-by-step
Ubutisi
URGENT PLEASE HELP. PLEASE MAKE SURE TO INCLUDE SCREENSHOTS WITH…

URGENT PLEASE HELP. PLEASE MAKE SURE TO INCLUDE SCREENSHOTS WITH EACH QUESTIONS ANSWER 

 

Use this link to complete the steps: https://playground.tensorflow.org/#activation=tanh&batchSize=10&dataset=circle&regDataset=reg-plane&learningRate=0.03&regularizationRate=0&noise=0&networkShape=4,2&seed=0.11892&showTestData=false&discretize=false&percTrainData=50&x=true&y=true&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=false&cosY=false&sinY=false&collectStats=false&problem=classification&initZero=false&hideText=false&batchSize_hide=true&showTestData_hide=true&discretize_hide=true&regularizationRate_hide=true&activation_hide=true&regularization_hide=true&problem_hide=true&stepButton_hide=true

 

Question 1:

-On the left hand side, where it says “Data,” there are four different arrangements of blue and orange dots to choose from. These are four different data sets that the network can learn to classify. Pick one of the four. Don’t change any of the other settings.

Take note of the background shading in the “Output” pane. What does it look like before training? Does it look like it will make a very accurate prediction?

Take note also of the connections between nodes. These are the initial, random connections. Which neurons are most strongly connected to which other neurons? it may be necessary to take a screenshot to be reminded of what the original configuration looked like.

-Now press the Play button, in the upper left, to begin training the neural network.

-You will see the blue and orange background in the “Output” pane, and the weights of connections between neurons, changing as the network begins to learn to make better predictions. You will also see a line plot that starts to move in the “Output” pane as the Test Loss and Training Loss numbers drop.

-When you think that the network has spent enough time learning, press Pause to stop the training.

How did the network change during this process? How did the predictions, the test and training loss, and the connections between neurons change? How do the patterns encoded in the hidden layer neurons change? Write the answers in a short paragraph, and attach an image of how the network looks now.

 

Question 2:

Now we are going to try the same thing again, but we are going to change some of the properties of how the network functions. Neural networks have a huge number of settings that can be changed based on the programmer’s goals.

Here are some parameters that you can change in the TensorFlow Playground:

Ratio of training data to test data: You can put more of the generated points into the training data or more into the test data. The network will only learn directly from points in the training data, so the more points are in the test data, the more challenging it will be for the network to generalize what it learns.

Noise:This is randomness that affects the data. With a high level of noise, orange and blue dots will be more scattered around the plot and may cross over each other’s “territories.” Neural networks are good at dealing with noisy data but the more noise there is, the harder they will have to work.

Features:These are the inputs that the network receives – they are fed into the first layer of the neural network in the form of numbers. By default, the network receives two types of input: the position of each dot on the X axis, and the position of each dot on the Y axis. However, you can also feed the network different types of numbers that are produced by performing mathematical operations on the first two. For instance, you can square the numbers before the network receives them, or multiply them by each other, or take their sines. You can include up to seven different input features in your network.

Number of hidden layers: By default, the number of hidden layers is two. You can reduce this to as few as zero, or increase it to as many as six. Each hidden layer is an additional set of connection weights and operations that the network can apply before producing its output.

Number of nodes in each hidden layer: By default, there are four nodes in the first hidden layer and two in the second, but you can change each layer to have any number of nodes between one and eight. You also don’t have to have the same number of nodes in each hidden layer.

Learning rate: This is the amount by which the neural network changes its connection weights every time it trains. A high learning rate means bigger changes to the connection weights will happen faster. A low learning rate means learning will be slower, but it might also capture more of the nuances of the data.

 

Choose at least two of these six parameters and change them in some way. You can also change to a different data set if you would like, although this does not count as one of your two required changes.

Write down which parameters you decided to change and why. Your goal should be, not just to get the “best” result, but to experiment with different possibilities of what can happen within a network while it learns.

Now do the same thing you did for Question 1 – press Play, watch the network learn, and press Pause when you feel like it’s finished learning.

What was different this time? How did your choices change how the network learned, in terms of accuracy (test and training loss), speed, or what happened to the connections between the nodes?

What do you think of these changes? What have you learned about the importance of these parameters for a neural network, and the different things that are possible when they are set up in different ways? If nothing important really seemed to change, why do you think that is?

Write down the answers in a paragraph the way you did for Question 1. Make sure to be specific about what was changed, and include a screenshot of what the network ended up looking like.