https://github.com/deep-diver/logistic-regression

simple neural network without hidden layer
https://github.com/deep-diver/logistic-regression

logistic-regression neural-network numpy

Last synced: 29 days ago
JSON representation

simple neural network without hidden layer

• Host: GitHub
• URL: https://github.com/deep-diver/logistic-regression
• Owner: deep-diver
• Created: 2018-04-02T07:03:09.000Z (almost 7 years ago)
• Default Branch: master
• Last Pushed: 2018-04-03T03:39:19.000Z (almost 7 years ago)
• Last Synced: 2024-10-04T13:42:18.893Z (3 months ago)
• Topics: logistic-regression, neural-network, numpy
• Language: Python
• Size: 5.19 MB
• Stars: 2
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Simple Neural Network without hidden layer

This repository is to demonstrate a simple neural networks without hidden layers. All the functions to build the neural networks are written in numpy. The simple model takes a batch of images of cat in shape shown below figure, and then it performs prediction.

![alt text](/images/capture1.png)

## The flow

### 1. **main** function

 - get the image datasets, then flatten and normalize them.

 - build and train the model

 - print out relative informations

### 2. **model** function

 - initialize parameters by calling **initialize_with_zeros** function

 - optimize the parameters by training on the training dataset. This is done by calling **optimize** function

 - run prediction on training and testing datasets by calling **predict** function

 - return cost, predictions, updated w and b, and etc.

### 3. **initialize_with_zeros** function

 - simply initialize the parameters, w and b

 - this is done by **numpy.zeros** function

### 4. **optimize** function

 - iterating for the number of iterations given as a hyper-parameter

 - while iterating it calculates gradient decent and costs by calling **propagate** function. also, update the parameters, w and b with the calculated gradient decent

 - return the final values of parameters and gradient decent

### 5. **propagate** function

 - performs calculation in the following order

  - **sigmoid** function as the activation function on w, X, and b ![alt text](/images/activatoin.PNG)

  - calculate the cost function J 
![alt text](./images/cost.PNG)

  - calculate the gradient decent on w and b

    - ![alt text](/images/dw.PNG)

    - ![alt text](/images/db.PNG)

    - ![alt text](/images/capture2.png)

return them

### 6. **predict** function

 - simply running **sigmoid** function on w, X, and b. with the result of the **sigomid** function (a), it returns 0 if a <= 0.5, and it returns 1 otherwise.

### 7. **sigmoid** function

 - simply calculate and return 
![alt text](./images/sigmoid.PNG)

## Dependencies

- **numpy**

  - for handy matrix operations

  - http://www.numpy.org/

- **h5py**

  - Pythonic interface to the HDF5 binary data format.

  - Image dataset is stored in h5 file format.

  - https://www.h5py.org/

## Reference

- Coursera Deeplearning.ai (by Andrew Ng.)

  - https://www.coursera.org/learn/neural-networks-deep-learning/notebook/Zh0CU/python-basics-with-numpy-optional