Skip to content

DataByteNITT/Inductions-22

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

45 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Inductions-22

Fork the repo and create a pull request of your solution to this repository for the corresponding tasks
Note : Do not send pull requests to the main repo. Make sure you are sending requests inside your tasks folder

Sentimental

medium link: https://medium.com/@advaith142001/farmers-protest-twitter-sentiment-analysis-a8aca1e52f43 ======= Parameter : A variable that is internal to the the model and whose value can be estimated from data. -They are required by the model when making predictions -They are often saved as part of the learned model

Hyperparameter : A variable that is external to the the model and whose value cannot be estimated from data. -They are often used in processes to help estimate model parameters. -They are often specified by the practitioner.

Gaussian process: It's a powerful algorithm for both regression and classification -Gaussian process is a probaility distribution over possible functions Kernal : The method of classifying linearly for the non-linear problems

Surrogate method : A statistical model to accurately approximate the simulation output

Probablistic model : Probabilistic modeling is a statistical approach that uses the effect of random occurrences or actions to forecast the possibility of future results -it provides a comprehensive understanding of the uncertainty associated with predictions. -Using this method, we can quickly determine how confident any mobile learning model is and how accurate its prediction is.

Nomenclatures:

  1. suurogate model (gaussian function in this case) It is the statistical/probabilistic modelling of the “blackbox” function. It works as a proxy to the later. For experimenting with different parameters This model is used to simulate function output instead of calling the actual costly function

  2. Acquisition Function It is a metric function which decides which parameter value that can return the optimal value from the function. There are many variations of it. We will work with the one “Expected Improvement”

Problem statement To summarize a research paper which talks about efficiency and implementation of bayesian optimaization

Pseudo code for Bayesian optimization

SURROGATE FUNCTION (Gaussian process) step1 Looping over all the samples values of input x, where the evaluatation takes place . 2. Building k and f vectors i.e the data 3. Building matrices X and Y 4. Calculating mu and sigma. 5. Appending mu to predictedMu array and sigma to predictedSigma array step6 Calculation of Omega as the mean of blackbox function for sampled points step7 Calculation of Kappa =PredictedMu + Omega step8 Returning values Kappa(estimated mean of suurogate func.) and predictedSigma (estimated variance of surrogate func.) I have used sklearn module to import gaussian Process in my model ACQUISITION FUNCTION Usually acquisition functiopn consist of : 1.Upper confidence bound 2.Lower confidence bound 3.Probability of imbprovement 4.Expected Improvement

Mathematical inpretation

let us take the actual function be f(x) bayesian function be y= f(x) + e(Etta) where e is small value to optimize the return value instead y can be represenated as gaussian distribution of (f(x), variance) GP is completely specified by its mean function mu(x) and its covariance k(x,x')

Loss function can be representated by gaussian distribution of its mean and covariance as mentioned above

Coming to acquisition function

expected improvement EI(x)= E[max {0, f(x)-f(x") where x" isthe current potimal set of hyperparameters. Maximizingthis parameter will give us the point that improves upon function the most

EI(x)= mu (x)- f(x"))psi(Z) + sigma (x)Pi(Z) if sigma (x) >0 = 0 if sigma (x) =0 therffore, Z= (mu (x)- f(x") )/sigma (x)

here, Psi (x) is cumulative function and pi(z) is probability density

Final points, 1.Given observed values f(x), update the posterior expectation of f using the GP model. 2.Find xnew that maximises the EI: xnew=argmaxEI(x). 3.Compute the value of f for the point xnew.

by iterating for different values we can make a perfect model or function which suits the actual functionmain

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published