Introduction | Learn Tech with Rookie

type

status

date

slug

summary

📝What is deep learning?

Deep learning is a subset of machine learning, which transform data(vedio text,etc.) to numbers and find patterns(unvisible relation or rules of generating output from input) of them.

📝What are neural networks??

There is a Input-Manipulation-Output structure, and the intermediate process is a bounch of layers(in each layer, there are many nodes connecting with nodes in last layers with different weights).

(I should put a picture here one day……)

📝Supervised and unsupervised learning and……

In supervised learning, we have data and labels. For example, we have 1,000 pictures of dogs and 1,000 pictures of cats. We tell the computer who those are dogs and who are cats like we are its superviser.

In unsupervised and self-supervised learning, we only give data to computer, and it does not know who is dog or cat.

In Transfer learning, we start a learning with algorithm already done. For example, we use the mode of distinguishing cats and dogs as an initial mode to a new algorithm(like distinguishing tiger and wolf).

Reinforcement learning(RL) is kind of like: we have a dog and teach it how to urine in toilet, when he pees on coach, we hit him and when he pee in toilet we give him a bone as reward.

DeepRL is that in a RL, we use neural network(or orther structure in DL) as the strategy we need to improve.

📝Concepts of tensor

Before talking about tensor, let us take a look on scalar, vector and matrix.

A scalar is basically a number like 5.6, which is used to represent some property, like weight, salary blabla

A vector is basically a seris of scalor, like(3.4, 53, 6), which is often used to represent some property with direction(e.g. coordinate in 3 dimention,like our living space, we can draw a line from(0,0,0)to(3.4, 53, 6)).

A matrix is basically seris of vector with the same dimension.

And all above are tensors, a tensor with order n is a seris of tensors with order (n-1). For example, vector is order-1, matrix is order-2, scalar is order-0. And dimension n1*n2*n3*… means the number of element in each layer.(maybe I should invovle a picture here)

We will see tensors in pytorch(torch.tensor) to see the basic block of machine learning.

📝Some tips using colab&Nvidia GPU

Colab is a great online web site for running .ipynb-format file which is developed by google.

In a .ipynb file, “Shift+Enter” is create a new code block. If one types “+Code”, a new text block shows up, which will not be compiled.

Runtime-change runtime type, follow clicking in this path, we can choose accelarating type(CPU/GPU/TPU). For deep learning, GPU is a good choice.

Typing this code can show the information of GPU.

观点2

引用的话语

🤗 总结归纳

总结文章的内容

📎 参考文章

一些引用

引用文章

💡

有关Notion安装或者使用上的问题，欢迎您在底部评论区留言，一起交流~