Regular Challenge

The problem consists in discriminating images representing deers from images representing horses.

The file Challenge_Train.mat contains the training set and is constituted by two matrices: Xtr and Ytr. Each row of Xtr is a vectorized 32x32 RGB image from the Cifar-10 dataset therefore it is a vector of 3072 elements whose first 1024 entries contain the red channel values, the next 1024 the green, and the final 1024 the blue. Ytr contains the labels.

Use the functions provided in the zip file of this challenge to find the algorithm and the right parameters that best solve the problem. In particular, the algorithm must be chosen among the following: k-Nearest Neighbors, Linear or Kernel Regularized Least Squares, Linear or Kernel Logistic Regression.

Once you find the right algorithm and parameters, produce a script that solve the problem for a generic test set Xts. Example of the script:

sigma = 0.1;

lambda = 0.02;

kernel = 'gaussian';

w = regularizedKernLSTrain(Xtr, Ytr, kernel, sigma, lambda);

Ypred = regularizedKernLSTest(w, Xtr, kernel, sigma, Xts);

Note that the parameters (sigma, lambda and the kernel in the example above) must be constants and the script must contain only the train and test functions of one of the algorithms specified above (in kNN they are condensed in one function).

Upload the script within the deadline on http://www.dropitto.me/alex86r

with password sendit. The name of the script must be FirstName_LastName.m

We will execute the script on our test set and compute the classification error, that is the score for the regular challenge (the smaller the better).

Wild Challenge (Optional, just for fun!)

Once you completed the previous part, you can do the wild part of the Challenge!!

The goal is to solve the same problem as in the regular challenge, but without the restriction of using only the functions provided in the zip file.

Here you are allowed to write your own learning pipeline. In particular you should to provide a function with signature

funtion Ypred = learningPipeline(Xtr, Ytr, Xts)

Within the deadline, upload a zip file, containing learningPipeline.m and all the functions and scripts needed for its execution, on http://www.dropitto.me/alex86r with password sendit. The name of the zipfile must be FirstName_LastName_WILD.zip

We will execute your learning pipeline on our test set and compute the classification error, that is the score for the wild challenge (the smaller the better).

NOTE: there is a time limit for the scripts. Every script that requires more than 5 minutes to output the predicted labels will be stopped.