At some point, perhaps not in class 9, it would be nice to know what are the weaknesses/limitations of CNNs–where are they struggling/weak/impractical?
I think that what you think about might be autoencoders.
Is video stream still cutting out for people? I need to keep reopening the link.
It did for few seconds again but was back immediately
cutting out for me
Is it possible to explain the concept of receptive field a bit more?
its offline now
@renjithmadhaven - try refreshing
From what I understand, the receptive field is the portion of the image used to compute the filter output.
Can we stick a CNN between this forum thread and Jeremy’s video stream?
Make it a dense one with no max pooling
Or Should we use Average pooling to retain the shape
video stream is down again
Down again!
Its working now.
It says offline
Just reopen it. Seems to be working now.
Still not working here…
Or we can use a generative model to generate this lecture from the course notes. Jeremy can work from home.