D(eep)L(earning)/practice{note, "2/7"}
ML differs from other coding
the key about machine learning is that we can generalize.
train a model with one set of data and apply it to a different set of data.
coding DL is harder than other forms of coding
always assume that every line of machine learning code you write is wrong (there's lots of ways to be wrong ), it's much harder to see that you're wrong.
problem: writing lots of code = = lots of things to make mistakes with + not have best practices in + hard to maintain.
in DL: less code is better.
DL in every domain just broke the state-of-the-art results in the field
4 key libraries are:
- numpy: numerical programming;
- matplotlib: plotting;
- pandas: tables of data;
- pytorch: deep learning;
remember: to make a leap forward you have to know how to use these fundamental libraries.
that would be enough to really give you all the basic knowledge you need.
practical approach 2 learning
it's important to be guided towards problem-solving.
in the reality of Deep Learning, you learn by doing and solving concrete problems.
it's not just about studying theory, but applying your knowledge in real situations, iterating and continuously improving.
DL is a field where you learn by doing.
it's not just about knowing the theory, but knowing how to solve concrete problems.
this means that even if you don't know all the mathematical details, you can start building models and understand the basics better as you progress.
you don't need to study years of linear algebra to start doing deep learning: most of the work boils down to simple matrix multiplications.
- learn as you go: focus on what you need to solve immediate problems.
- don't worry about understanding everything right away.
over time, you'll gain a deeper understanding of the concepts. - experimenting directly with models is the key to developing a practical understanding.
in-depth(theory+knowledge):
focus first on the practical learning of how to train and use neural networks, and only then delve into the underlying mathematics.
you don't need to know all the theoretical details to start getting results.
learn the why and how of things as you need them to move to the next level.