A 2D line is defined by two properties:
1) A y-intercept (called a "bias" in this context)
2) A gradient (called a "weight" in this context)
These values are (respectively) added to or multiplied by the rest of the function to produce our final line.
So, if we want to change the line we fit to our graph, we need to change our weights and biases.
Weights and biases are examples of "parameters", and technically speaking, there can be far more than two for any given function.
But to avoid the complexity of drawing million-dimensional graphs, we're going to stick with linear functions.
For now, just understand that, to improve the accuracy of a regression model, we need to tweak the parameters that make up the line on our graph.
And to understand how, we use cost functions.