Retrospection of Svm Classifier

SVM (Support Vector Machine) is a supervised learning which is a boon in disguise to the field of machine learning. Though a number of classifier seems to exist it gives better result and recognition rates for which it is opted the most. The other brighter side of SVM is that it minimizes the empirical error and maximizes the geometric region. Neural network has weakness such that they converge only to the locally best solutions. Whereas, SVM is far improved. SVM has the capability to select its own support vectors. In case of Back Propagation algorithm, we should know in advance the value of the output and once we receive a value after passing through the neurons, the two values are compared and if there is no match found, backtracking is done as a result of which weights are varied to obtain the exact value. The computational complexity is going to be very great. It proves to be a useful tool when the data is not regular or when the distribution is unknown. SVM gains its flexibility from the kernel which in turn makes it successful. This can provide a unique solution whereas neural networks have multiple solutions for each minima so it does not seem to be robust for different samples. The greater recognition rate and flexibility makes SVM popular. Accuracy, recognition rate is very important for the purpose of classification, only then authentication can be done effectively. This study has been evolved to reveal that SVM gives good accuracy and recognition rate compared to other classifiers and hence it is considered best for gait recognition.