- This topic has 1 reply, 2 voices, and was last updated 8 years ago by
.
- Posts
Hi there
Thank you for the fabulous lectures – and can’t believe they are free, how wonderful!
I struggle with one part of the lecture…towards the end there he works out the learning rate given the data for 16 batches.
Firstly,I don’t understand : Is he using the formula y=a.x^b ?
By “r” and by “learning rate” does he mean ^b?
i.e. is the learning coefficient the same as the learning rate ?Or is he busy with the first method, doubling, where one multiplies the average time p.u of the previous unit tiby the learning rate (e.g. 85% or 0.85) to get the average time p.u of the next unit.
Thank you to John for his lovely videos and all the hard work put into this platform.
Kind regards
Lise GreylingHi – thanks for your feedback.
The figure ^b is known as the learning coefficient. This means you will use it as an index figure – ‘to the power of’ in the learning curve formula.
The learning coefficient is calculated as: log r / log 2
where ‘r’ is the learning rate expressed as a decimalFor example if the learning rate is 80% then the coefficient figure ^b will be log0.8/log2.
If entered into calculator correctly this should give figure -0.3219 (to 4 decimal places)In terms of the last technique John is showing how to you can find the learning rate when this is not provided in the question.
This technique refers back to the table method – and so will only be required when units made are relatively low numbers. He shows how you can calculate the rate by counting the number of jumps between the first unit made and the most recent unit’s average time. Then by square rooting to the power of this figure we can find the value of the learning rate ‘r’ (given as a decimal).
Please also view this chapter in the CIMA P2 notes of Open Tuition resources this gives the calculated example with full answers at the back of notes ( see exercise 5 – chapter 5)
Many thanks
- You must be logged in to reply to this topic.