Evaluation of programs codes using machine learning
Download as PPTX, PDF0 likes34 views
This document discusses using machine learning to detect copied code submissions. It proposes using unsupervised learning via k-means clustering and dimensionality reduction with principal component analysis (PCA) to group similar codes and reduce complexity from O(n^2) to O(n). Key steps include extracting features from codes, applying PCA to reduce dimensions, running k-means to cluster codes, and detecting copies between clusters. This approach could help identify cheating in online programming contests and evaluate student code submissions.
![ALGORITHM
• Reducing data from n-dimensional to k-dimensional
• Compute the covariance matrix
•This is commonly denoted as
•Σ (greek upper case sigma) - NOT summation symbol
•Σ = sigma
•This is an [n x n] matrix
•Remember than xi is a [n x 1] matrix
•In MATLAB or octave we can implement this as follows;](https://image.slidesharecdn.com/evaluationofprogramscodesusingmachinelearning-200112005756/85/Evaluation-of-programs-codes-using-machine-learning-16-320.jpg)
![•Compute eigenvectors of matrix Σ
•[U,S,V] = svd(sigma)
•svd = singular value decomposition
•More numerically stable than eig
•eig = also gives eigenvector
•U,S and V are matrices
•U matrix is also an [n x n] matrix
•Turns out the columns of U are the u vectors we want!
•So to reduce a system from n-dimensions to k-dimensions
•Just take the first k-vectors from U (first k columns)](https://image.slidesharecdn.com/evaluationofprogramscodesusingmachinelearning-200112005756/85/Evaluation-of-programs-codes-using-machine-learning-17-320.jpg)
Evaluation of programs codes using machine learning
- 1. EVALUATION OF PROGRAMS CODES USING MACHINE LEARNING SUBMITTED BY – MANAS CHHABRA 2K12/SE/041 ROHIT PAL 2K12/SE/ 066 TANMAY AGGARWAL 2K12/SE/
- 2. OBJECTIVE •To detect copied codes submitted by different users on online judges.
- 3. HOW CAN WE SAY WHETHER TWO CODES ARE COPIED ? • -Tradition Way • Make a program which compare every code submitted by user to every other code in the database.
- 4. THEN WHAT IS THE PROBLEM WITH THIS METHOD Assume that we have 4000 user to solve a given programming question, assume that every code has 30 lines of codes
- 5. WE CAN RESOLVE THIS COMPLEXITY USING MACHINE LEARNING And Before we know how we have reduced this complexity Let’s first Know something about the machine learning concepts which we have applied
- 6. WHAT IS MACHINE LEARNING? • According to Arthur Samule , Machine Learning is a Field of study that gives computers the ability to learn without being explicitly programmed. • According to Tom Mitchell it is a Well-posed Learning Problem: A computer program is said to learn from experience E with respect to some task T and some performance measure P , if its Performance on T , as measured by P, improves with Experience E.
- 7. MACHINE LEARNING ALGORITHMS • Supervised Learning • Unsupervised Learning
- 8. WE ARE GOING TO USE UNSUPERVISED MACHINE LEARNING Clustering : Learning from unlabled data. Unsupervised learning Try and determining structure in the data Clustering algorithm groups data together based on data features
- 9. WHAT IS CLUSTERING GOOD FOR • Market segmentation - group customers into different market segments • Social network analysis - Facebook "smartlists" • Organizing computer clusters and data centers for network layout & location • Astronomical data analysis - Understanding galaxy formation
- 10. K-MEANS ALGORITHM • Used to automatically group the data into coherent clusters • e.g. Assume for this unlabled data • Step 1Randomly allocate two points as the cluster centroidsCluster • Step 2 Go through each example and depending on if it's closer to the red or blue centroid assign each point to one of the two clusters
- 11. • Step 3Move centroid step • Take each centroid and move to the average of the correspondingly assigned data-points • Repeat Step 2 and Step 3 until convergence
- 12. MORE FORMAL DEFINITION • INPUT: • K (number of clusters in the data) • Training set {x1, x2, x3 ..., xn) • Algorithm • Randomly initialize K cluster centroids as {μ1, μ2, μ3 ... μK}
- 13. DIMENSIONALITY REDUCTION • Speeds up algorithms • Reduces space used by data for them • Reduce dimension from nD to mD e.g. Reduction of 3D->2D
- 14. PRINCIPLE COMPONENT ANALYSIS (PCA) • To reduce from nD to kD weFind k vectors (u(1), u(2), ... u(k)) onto which to project the data to minimize the projection error • So lots of vectors onto which we project the data • Find a set of vectors which we project the data onto the linear subspace spanned by that set of vectors • We can define a point in a plane with k vectors
- 15. • e.g. 3D->2DFind pair of vectors which define a 2D plane (surface) onto which you're going to project your data • Much like the "shallow box" example in compression, we're trying to create the shallowest box possible (by defining two of it's three dimensions, so the box' depth is minimized)
- 16. ALGORITHM • Reducing data from n-dimensional to k-dimensional • Compute the covariance matrix •This is commonly denoted as •Σ (greek upper case sigma) - NOT summation symbol •Σ = sigma •This is an [n x n] matrix •Remember than xi is a [n x 1] matrix •In MATLAB or octave we can implement this as follows;
- 17. •Compute eigenvectors of matrix Σ •[U,S,V] = svd(sigma) •svd = singular value decomposition •More numerically stable than eig •eig = also gives eigenvector •U,S and V are matrices •U matrix is also an [n x n] matrix •Turns out the columns of U are the u vectors we want! •So to reduce a system from n-dimensions to k-dimensions •Just take the first k-vectors from U (first k columns)
- 18. • Now if we need to reduce to k dimension • Then we extract k columns from matrix U to Ureduce
- 19. NOW LET’S IMPLEMENT THIS INFORMATION TO OUR PROJECT
- 20. FEATURES TO BE CONSIDERED - Detect type of file eg . .java, .cpp etc - No. of Lines of Codes -No. of Functions -No. of Variables used -No. of if-else conditions -No. of loops
- 21. APPLICATIONS • Beside detection of Cheating in Online programming contests we can have following applications • We can use this in our Programming labs to evaluate the programs submitted by students according to complexity. • We can know what programming style is in trend these days
- 22. THANK YOU !