For work, I write down everything I want to do each day and then check off the things I actually did. Keeping tasks bite-sized, like an hour or less, increases the chance atask can be completed. Finally, when its time for your parents to decide your fate, you can share the notebook, so that even if you didn't reach all of their demands, you can demonstrate you now are at least organized and moving in the right direction.

Since CNN weights are often 4D tensors, WxHx(#inputs)×(#outputs), they're hard to visualize directly. Instead, the trick is to ask what input images to the model fully activate the queried node and none of the other nodes in the same layer. There's a Keras example script that does this. The generic term is "deep dream."

How badly the editor wanted to say rates "skyrocketed"?

Based on a cursory Google search of your concern, it looks like the manufacture, transport, and installation cancels out the 1st year of benefits. However, the next 19-24 years should be "clean." Oh, but then recycling will a new challenge. Wake me in 25 years.