Scribd
Upload
634 views3 pages

Code C# Adaboost

This document discusses using AdaBoost to perform multi-class classification on letter recognition data. It loads letter data, unrolls the data to transform the multi-class problem into multiple binary classification problems, trains an AdaBoost classifier on the unrolled data, then evaluates the classifier's performance on train and test data.

Uploaded by

Lu Bu Lu Ba
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
634 views3 pages

Code C# Adaboost

This document discusses using AdaBoost to perform multi-class classification on letter recognition data. It loads letter data, unrolls the data to transform the multi-class problem into multiple binary classification problems, trains an AdaBoost classifier on the unrolled data, then evaluates the classifier's performance on train and test data.

Uploaded by

Lu Bu Lu Ba
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 3

using System; using System.Diagnostics; using using using using using Emgu.CV; Emgu.CV.Structure; Emgu.CV.ML; Emgu.CV.ML.Structure; NUnit.

Framework;

namespace AdaBoostTest { [TestFixture] public class AdaBoostLetterTest { private static void ReadLetterData(out Matrix<float> data, out Matrix<fl oat> response) { string[] rows = System.IO.File.ReadAllLines("letter-recognition_data .data"); int varCount = rows[0].Split(',').Length - 1; data = new Matrix<float>(rows.Length, varCount); response = new Matrix<float>(rows.Length, 1); int count = 0; foreach (string row in rows) { string[] values = row.Split(','); Char c = Convert.ToChar(values[0]); response[count, 0] = Convert.ToInt32(c); for (int i = 1; i < values.Length; i++) data[count, i - 1] = Convert.ToByte(values[i]); count++; } } [Test] public void LetterTest() { Matrix<float> data, responses; ReadLetterData(out data, out responses); Debug.WriteLine("Data is loaded"); const int class_count = 26; int nsamples_all = data.Rows; int ntrain_samples = (int)(nsamples_all * 0.5); int var_count = data.Cols; // Create boost classifier // // // // // // // // !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! As currently boosted tree classifier in MLL can only be trained for 2-class problems, we transform the training database by "unrolling" each training sample as many times as the number of classes (26) that we have. !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

// 1. unroll the database type mask Debug.WriteLine("Unrolling the database..."); Matrix<float> new_data = new Matrix<float>(ntrain_samples * class_co unt, var_count + 1); Matrix<float> new_responses = new Matrix<float>(ntrain_samples * cla ss_count, 1); for (int i = 0; i < ntrain_samples; i++) { for (int j = 0; j < class_count; j++) { int new_row = i*class_count + j; for (int k = 0; k < var_count; k++) { new_data[new_row, k] = data[i, k]; } new_data[new_row, var_count] = (float)j; new_responses[new_row, 0] = (float)Convert.ToInt16(responses [i, 0] == j + 'A'); } } // 2. create type mask Matrix<byte> var_type = new Matrix<byte>(var_count + 2, 1); for (int i = 0; i < var_count; i++) { var_type[i,0] = (byte) Emgu.CV.ML.MlEnum.VAR_TYPE.NUMERICAL; } //// the last indicator variable, as well //// as the new (binary) response are categorical var_type[var_count, 0] = (byte)Emgu.CV.ML.MlEnum.VAR_TYPE.CATEGORICA L; var_type[var_count + 1, 0] = (byte)Emgu.CV.ML.MlEnum.VAR_TYPE.CATEGO RICAL; // 3. train classifier Debug.WriteLine("Training the classifier (may take a few minutes)... "); MCvBoostParams boostParams = new MCvBoostParams(); boostParams.boostType = Emgu.CV.ML.MlEnum.BOOST_TYPE.REAL; boostParams.weakCount = 100; boostParams.weightTrimRate = 0.95; // Note : to turn off set to 0 boostParams.maxDepth = 5; boostParams.useSurrogates = false; boostParams.maxCategories = 2; Boost boost = new Boost(); boost.Train(new_data, Emgu.CV.ML.MlEnum.DATA_LAYOUT_TYPE.ROW_SAMPLE, new_responses, null, null, var_type, null, boostParams, true);

Debug.WriteLine("Training complete"); Matrix<float> temp_sample = new Matrix<float>(1, var_count + 1); Matrix<float> weak_responses = new Matrix<float>(1, boostParams.weak Count); double train_hr = 0; double test_hr = 0; // compute prediction error on train and test data for( int i = 0; i < nsamples_all; i++ ) { int best_class = 0; double max_sum = -double.MaxValue; double r; for(int k = 0; k < var_count; k++ ) temp_sample[0,k] = data[i,k]; for(int j = 0; j < class_count; j++ ) { temp_sample[0,var_count] = (float)j; boost.Predict(temp_sample, null, weak_responses, MCvSlice.Wh oleSeq, false); double sum = CvInvoke.cvSum(weak_responses).v0; if( max_sum < sum ) { max_sum = sum; best_class = j + 'A'; } } r = Math.Abs(best_class - responses[i,0]) < float.Epsilon ? 1 : 0; if( i < ntrain_samples ) train_hr += r; else test_hr += r; } test_hr /= (double)(nsamples_all-ntrain_samples); train_hr /= (double)ntrain_samples; Debug.WriteLine("Recognition rate: train = " + train_hr*100 + ", tes t = " + test_hr*100); Assert.AreEqual(82.06, train_hr*100); Assert.AreEqual(78.43, test_hr*100); Debug.WriteLine( "Number of trees: " + boostParams.weakCount); Assert.AreEqual(100, boostParams.weakCount); } } }

You might also like