BLOG main image
분류 전체보기 (658)
소프트웨어 개발 (284)
크리스찬 (35)
일반 (303)
자녀교육 (15)
문서, 팁 (3)
나의책읽기 (15)
180,252 Visitors up to today!
Today 6 hit, Yesterday 3 hit
daisy rss
2018.04.03 14:56


배포 유틸리티는 파이썬에서 모듈을 빌드하고, 패키지를 만들고, 배포할 수 있게 해 준다. 

일단 배포 패키지가 만들어지면 모듈을 PyPI에 올려서 전 세계에 공유할 수 있을 뿐만 아니라 여러분 컴퓨터에도 모듈을 설치할 수 있다. 

1. 폴더를 만들고 모듈 파일 포함

2. 폴더에 파일 생성

이 파일은 배포 패키지에 대한 메타데이터를 갖는다. 

3. 배포 패키지 생성

python3 sdist

4. 배포 패키지 설치 

python3 install

PyPI 웹사이트에 등록하기 

1. http://pypi.python.org에 가입

2. python3 register로 등록 

3. python3 sdist upload로 패키지 업로드  

2018.04.03 13:09

온라인 세미나로 진행했습니다.

세미나 진행하면서 요약한 내용은 다음 블로그에서 볼 수 있습니다.

2018.02.18 22:54

Major Architectures of Deep Networks

Unsupervised Pretrained Networks (UPNs)

- Autoencoders

- Deep Belief Networks (DBNs)

- Generative Adversarial Networks (GANs)

Convolutional Neural Networks (CNNs)

Recurrent Neural Networks

Recursive Neural Networks

CNN 아키텍처는 Input layer - Feature-extraction (learning) layers - Classification layers로 구성된다. 

The feature-extraction layers have a general repeating pattern of the sequence: Convolution layer, Pooling layer

Recurrent Neural Networks are considered Turing complete and can simulate arbitrary programs (with weights). Recurrent neural networks model the time aspect of data by creating cycles in
the network. A well-trained Recurrent Neural Network could compete in Alan Turing’s famed Turing Test, for instance, which attempts to fool a human into thinking he is speaking with a real person.

Recurrent Neural Networks are trained to generate sequences, in which the output at each time-step is based on both the current input and the input at all previous time steps. Normal Recurrent Neural Networks compute a gradient with an algorithm called backpropagation through time (BPTT). Recurrent Neural Networks change this input dynamic to include multiple input vectors,
one for each time-step, and each vector can have many columns.

LSTM networks are the most commonly used variation of Recurrent Neural Networks. LSTMs are known for the following: Better update equations, Better backpropagation