Skip to content

Latest commit

 

History

History
111 lines (82 loc) · 3.4 KB

README.md

File metadata and controls

111 lines (82 loc) · 3.4 KB
Raw

Infectious Disease Models

Python implementation of common infectious disease transmission models.

A Gallery of infectious disease models for the quick usage in the infectious disease control and prevention.

Cengceng is a name of a cat my son loves.

quick start

1. Installation

This project use pytorch as the backend mathmatical framework, a pre-installed GPU version of pytorch is recommended.

pip install cengceng

2. Usage

  1. Project predefined model
from cengceng.compartmental import Sir

model = Sir()

# load data from csv file
model.load_data('./covid-19-cn.csv')
model.fit()
model.save('output/result.csv')

That's it.

  1. Customize your own model
from cengceng.models import Model

# Inherit from basic model class
class My_Own_Model(Model):
		def __init__(self):
				super().__init__()

Models:

Compartmental Models

These types of models seperate each individual into states. And the speed of one state to another was predefined. In cengceng, we define the state change function in the forward() function of model class. Models included:

  • SIR
  • SEIR
  • mosquito resistance prediction

For instance, in the simplest SIR model:

class SIR(Model):
		# defined in the forward function
    def forward(self, s0, i0, r0):
        si = s0
        ii = i0
        ri = r0
        dsdt = -self.beta * si * ii
        didt = self.beta * si * ii - self.gama * ii
        drdt = self.gama * ii
        return torch.cat((dsdt, didt, drdt))

Definitive Models

Models with definitive function

  • logistic

Modified Infectious Disease Models

  • seir-caq

Malaria Transmission Model

States

At each point people can be in one of six states:

  • S - susceptible
  • T - treated clinical disease
  • D - untreated clinical disease
  • A - asymptomatic infection(which maybe detected by microscopy)
  • U - sub-patent infection
  • P - protected by a period of prophylaxis from prior treatment

Development

Skip this part if you are not intended to contribute to the project

But it will be helpful, if you understand the interior mechanism of how this projcet works

Create Your Own Model

  1. Compartmental Model Inherite from Compartmental Class, useadd_trainable_parameter function to add trainable parameters.
class My_Compartmental_Model(Compartmental):
		def __init__(self, para1=1e-5):
				super().__init__()
				self.add_trainable_parameter('para1', para1)

Compartmental Model's Implementations

This project use pytorch as a major mathmatical framework to calculus and differentiation. The parameters are all basically a torch variable. Because models including sir, seir and other compartmental models can't directedly calculate the original function, the project use torchdiffeq to calculate the ode function.

RNN Models

Recurrent Neural Network(RNN) is a class of recurrent neural network (RNN) is a class of artificial neural network where connections between nodes form a directed graph along a temporal sequence. This allows it to exhibit temporal dynamic behavior. Unlike feedforward neural networks, RNNs can use their internal state (memory) to process sequences of inputs. This makes them applicable to tasks such as unsegmented, connected handwriting recognition or speech recognition.

Model Training

References

All references are stored in the references folder and was stored as .bib type.