How do you measure how big a convolutional neural network is?
You can't weigh it or use a ruler to measure it. And if you can't measure it…then how can you scale it? Until 2020, the process of measuring a convolutional neural network was never well understood. That is until researchers set out to answer an important question:
Is there a principled method to scale up ConvNets, so they achieve better accuracy and efficiency?
And in the process, they accomplished two feats which changed the direction of deep learning:
1) Discovered a novel scaling method called compound scaling.
2) Created a new family of SOTA architectures called EfficientNet.
Now, back to the original question: how do we measure the size of a ConvNet?
By looking at three factors:
1) Resolution (dimensions of its inputs)
2) Width (number of feature maps)
3) Depth (number of layers in the network)
All three factors — depth, width, and resolution — impact the accuracy and efficiency of your network. Ideally, you want to maximize all these factors and accomplish the following:
• Retain the baseline model architecture, i.e. keep the operations in each layer fixed.
• Leave the memory footprint of your model constrained to some target hardware.
• Keep the number of FLOPs below some predefined threshold.
But there's a catch…
Scaling up only one network dimension (width, depth or resolution) improves accuracy, but the accuracy rapidly diminishes. For better accuracy and efficiency, you must balance all network width, depth, and resolution dimensions during ConvNet scaling.
Coupons are issued by instructors to promote their courses, gain traction and reach momentum. The instructor can choose to emit discounted (ex: $11.99 coupon) or 100% off coupon (you pay nothing). Each coupon becomes expired when emitted quota is over (1000 enrollments) OR expiration date has been reach (5 days).
For a coupon, number of activation are now capped to 1000 max. This means that it can be activated only a 1000 times, and then it expires; or reach its expiration date; whatever happens first.
We have no contact with instructors, and only instructors can emit coupons. You can try to directly contact the instructor finding his/her Twitter/Facebook, and ask him/her for a coupon, but at our level, we cannot help, sorry.
We have an affiliate contract with Udemy and we may receive a commission when you purchase through some of the affiliate links on this website. But this website is not a part of the Udemy Inc. Additionally, this website is NOT endorsed by Udemy in any way. Udemy is a trademark of Udemy, Inc. `