If α controls the speed of learning, how about increasing it to learn faster? Try using α = 0.5 in the previous calculations. What a and b would you get? Try to sketch a where the line of the new model would be on the graph in relation to the data. What were the consequences of increasing α?

Question

Answer

The model starts to diverge from data because the step of gradient descent is too large.

Answer

Learning is faster, but otherwise no difference

Answer

The learning became slower but on the positive side it is very stable now.

Answer

The model was perfectly optimized in one step, should have used α = 0.5 in the first place

Crowdly

If α controls the speed of learning, how about increasing it to learn faster? Tr...

Want instant access to all verified answers on moodle.taltech.ee?