richadutt Goto Github PK

bitcoin-simulator

Implementation of Bitcoin protocol to simulate bitcoin mining, wallet, and transactions.

job-scheduler-using-red-black-and-min-heap.-

linear-and-circular-convolution-using-fft

This program demonstrates (i) the speedup obtained by using FFTs in numerical convolution. The two sequences x and y must contain at least 1000 elements each. The convolution code is written on own and libraries are used for the FFT computation. The speedup is documented using TIC TOC. (ii) The errors between circular convolution using FFTs and linear convolution (direct computation) is documented. In both (i) and (ii), 5 sets of random x and y sequences are used.

music-genre-classification

Classify music in two categories progressive rock and non-progressive rock using mfcc features, MLP, and CNN.

principal-component-analysis

An octave script to generate the principal components of the clockwork-angels. This is run for different choices of input parameters. The relevant ones are: (i) Number of patches (number_patches) chosen to be somewhat greater than 1000. The code checks for this and discards duplicates. We also present a result for 20000 patches; (ii) Patch size (patch) usually chosen to be 16 × 16 but can be increased (and we present a result for 24 × 24); (iii) The number of eigenvectors (number_eig) chosen for the final display. This is usually 64 but we present a result with 256 eigenvectors; (iv) The gap between the eigenvector images (scratch) for the final display usually set to 4. There are many possible criteria for deciding how much information is preserved. One of the best is the sum of the Frobenius norm errors of all reconstructed patches divided by the Frobenius norm of the patch. But, the ratio of the sum of chosen eigenvalues divided by the total sum of all eigenvalues is also reasonable, etc

richadutt

svd-based-image-reconstruction

Load the hendrix_final.png image and extract the R, G and B channels. Convert each channel image to double precision. Then execute the SVD separately on the R, G and B channels of the image. Plot (using a log-log plot) the non-zero singular values for the R channel. Comment on the nature of the plot. Plot the Frobenius norm of the reconstruction error matrix for each channel w.r.t. the dimension (increasing from 1 to the rank) and display the original and final reconstructed images (combined from R, G and B reconstructions)