I have been following the lectures so far, and everything is understood very well.
What problem I am facing is that I find few parts of the code a bit difficult to understand and work my way through them.
Can you please suggest some ways in which I can improve and overcome this
You can try this book to understand coding.
Python Data Science Handbook by Jake
Vanderplas. If you couldn’t find it, then mention your mail id here I’ll send you a copy. I hope it helps.
Thanks in advance xD
I went through the book and i guess my doubt misinterpreted.
I would like to provide a snippet as an example of code I am having a hard time with
it would be helpful if you provide with insights referring above pictures
I have made few notes while understanding the code and the visualisation. I am sharing it so that it might help you.
Function meshgrid -
- meshgrid is used when we have two different axis and we want to look at every possible combination of the two and then compute them.
- Contour plots (sometimes called Level Plots ) are a way to show a three-dimensional surface on a two-dimensional plane. It graphs two predictor variables X Y on the y-axis and a response variable Z as contours. In this case x and y are the two features and z is the response you see in the form of dots represented corresponding to each point with its feature colour. However we can further introduce 4th dimension by varying the size of the dots as discussed in the videos.
Coming to the explanation of plots drawn(visualisation)-
- The three contour plots we see are representing the three nuerons h1, h2, h3 repectively.
- In the plots above it is clear how h1 is dividing a boundary
- then in second plot it is dividing anonther boundary with the second neuron h2
- And then finally in the third plot both the boundaries can be seen through as the prediction passes through h3
- Also if we see closely we can notice after passing through each neuron the boundary becomes thinner and more accurate.
Heat maps -
- Heat maps are represeting the values of the different weights and biases.
- Negative values represented by red colour. Values close to zero are represented by yelllow colour. Positive values are represented by green colour.
- Please refer to the image for structural visiualisation of nuerons in each layer.
- Index 11,12,13,14,02,03 are weight terms
- Index 10, 15, 01 are the biases.
- Why do these specific indices represent weight and bias ?
- Beacuse this is how we created our weight_matrix and bias matrix. Refer to the image for structural intuition.
- Another thing to notice from the heat map is there is significant value of each weights this means every feature is necessary and responisble in computing the true output whereas if in case few of the weights were close to zero this would indicate that it is not contributing much which in turns means that those input features are not contributing much and hence not much important so we can actually drop them. This is the main purpose of drawing the heat maps here.
Feel free to correct me as these are my interpretation to the explanations in the video.
Sorry, I’m still learning 10th week. Since your post mentioned codes I had misinterpreted as general level coding.