Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 5

Deep learning

CHAN YING PUI 2023318041


Feedforward Neural Network
- a fully connected network which generate result with layers

- multi-layered network of neurons

- the information only past forward

- the process: receive data through hidden layers export output

- specially used in classification & regression

- more data= more accuracy


AutoEncoders
- unsupervised algorism

- mainly constructed with encoder and decoder

- reconstruct the data to possibly generate an output that is equivalent to the input

- include the process of compressing input data, encoding them and reconstruct the data as an output

- help to reduce the complexity of input data, we do not need a lot of specific values to generate an output

- systematically reduce the complexity by the multiple layers into compressed version
Convolutional Neural Network
- learn directly from data

- finding patterns to identify objects and categories

- could be conveniently used for classifications

- e.g., face recognition

- can automatically detect important values and features without human supervision

- reduces the high dimensionality of images without losing its information


Self-learning & self-aware
- train itself using unlabeled data

- even if the data input are not structured detailly by helps, they can generated data labels automatically with
an output

- they could analyze the dataset and export output by searching for the the patterns and matches with a
results, which could be simply by training itself with more dataset that are input beforehand

You might also like