rk68 (5) [Avatar] Offline
#1

- Re fig. 1.4 and 1.5: I would add an arrow and 'Prediction' to fig. 1.5. Now it looks like 'Model' is the end product of the inference (as it is for the learning). I think this is confusing.

- Re fig. 1.9: I would add 'Model' to the top part of the picture as well. The model is created as part of the training process, so should be part of the top box. It is then applied (as is shown) in the bottom box.

- Listing 2.1 doesn't give you the dot product. It gives c1*a1+c1*a2+c1*a3+.....+c3*a2+c3*a3 instead of c1*a1+c2*a2+c3*a3 for costs=[c1,c2,c3] and amounts=[a1,a2,a3]
Nishant Shukla (52) [Avatar] Offline
#2
Hi rk68,

Thanks for your feedback!

rk68 wrote:
- Re fig. 1.4 and 1.5: I would add an arrow and 'Prediction' to fig. 1.5. Now it looks like 'Model' is the end product of the inference (as it is for the learning). I think this is confusing.

I agree. It's more clear to label it 'Prediction' since the final produce of learning is to create a 'Model.'


- Re fig. 1.9: I would add 'Model' to the top part of the picture as well. The model is created as part of the training process, so should be part of the top box. It is then applied (as is shown) in the bottom box.

Correct. I think I'll fix this by reusing hte model figure in both the learning and inference stages.


- Listing 2.1 doesn't give you the dot product. It gives c1*a1+c1*a2+c1*a3+.....+c3*a2+c3*a3 instead of c1*a1+c2*a2+c3*a3 for costs=[c1,c2,c3] and amounts=[a1,a2,a3]

Good catch! Fixing immediately.




22152 (1) [Avatar] Offline
#3
Hi,
some more comments,
§1.1 young technology, but existing since 80's, 90's see ICML edition number
§1.7.1 lib , you can quote Keras and Lasagne above Theano. Note Keras support now 2 backends = Tf and Theano
pandas to be included in famous lib.

Thanks for this book.!

Best Regards
Bruno
425150 (1) [Avatar] Offline
#4
I bought the MEAP yesterday and listing 2.1 is still
revenue = 0
 for cost in costs:
     for amount in amounts:
         revenue += cost * amount

still not the dot product of vectors
Nishant Shukla (52) [Avatar] Offline
#5
Hi 425150,

Sorry about the delay. I'm revising chapters 1-6 together in a major bulk update.

Almost done!