Yeshwanth

Hypothesis for experiment 4 :

For a given training data size, the accuracy of detecting a circle is more than the accuracy of detecting a wave

Testing : The hypothesis can be tested by training two models for the corresponding symbols for a fixed training size. Accuracy of both the models can be calculated and compared.

Hypothesis: SK Algorithm try to classifies the correct image of letter from training data Test: From the test data, identify the correct test data base on the training data (image files)

Hypothesis 1: Accuracy of SVM model may be improved by increasing the training data. Testing : Increase training data in increments and observe the svm convergence. Hypothesis 2. Performance of model doesn’t change when identifying different images.

1. Increasing the number of training examples would increase the accuracy of the model. 2. For the same number of training examples learning Complex symbols is less accurate than that of simpler ones. 3. Increasing the noise would reduce the accuracy

Hypothesis: Accuracy will increase as size of training set increases. Testing: Compare the accuracy with different sizes of training sets.

Name: Abhinaya Koduri Hypothesis 1: Accuracy may be improved by increasing the training data size. Testing : Increase training data in increments and observe the svm convergence. Hypothesis 2: The model will be dependent on the quality of train data and the complexity of the symbols. Complex ones take more time Testing: Train and test the model with simpler symbols and then with complex ones

Hypothesis: Number of examples choosen for training examples is propotional to the accuracy achieved by the training model. Hypothesis 2: To train complex images, the number of training examples should be more.

Hypothesis : We can improve the accuracy of SVM by increasing the size of training set.

Testing : We will train SVM on different sizes of training set, and calculate the accuracy of each model. We will then plot each size with the corresponding accuracies. For our hypothesis to be correct, we should see a upward trend.

Hypothesis: Requires greater number of training examples to learn complex symbols as opposed to learning simpler symbols

Testing: This can be tested with an SVM as bi-classifier. A model can be built for each symbol and accuracy calculated for each relevant case.

Hence, simpler symbols achieve higher accuracy by using fewer training examples.

Hypothesis: A large number of training samples results in better prediction for SVM. Test: Increase the number of training samples and compare accuracy