Public discourse on AI systems tends to be centered around machine-learning approaches. One of the dominant techniques for machine learning is that of neural networks in which layers of perceptrons (see the previous module) are stacked one after another to solve problems that cannot be satisfactorily solved with a simple combination of linear separations of the inputs.
We could of course build such a system from simple perceptrons, feeding the outputs of the first layer into the second one as inputs and so on, but the modern-day approach is less artisanal: deep learning refers to combining layers with different, predefined internal workings (a bit like LEGO bricks) to build a pipeline of sorts.
This module will help you do the following:
Spending a few minutes before live class with the Baheti's The Essential Guide to Neural Network Architectures will help make sense of what we discuss.
After having glanced at the above blog, reflecting upon the concept of architecture with respect to buildings (as well as that of computer architecture or software architecture, if you are familiar with it at all), elaborate a bit in your own words what the elements do you precieve the concept of neural network architecture to entail as components. What other words can you think of, if any, that might describe the overall concept as well as or better than "architecture"?
After this module, you should be familiar with the following concepts:
Remember that you can always look concepts up in the glossary. Should anything be missing or insufficient, please report it.