2017-10-11

Oct 11 In-Class Exercise.

Post solutions to the Oct 11 In-Class Exercise to this thread.
Best,
Chris
Post solutions to the Oct 11 In-Class Exercise to this thread. Best, Chris

-- Oct 11 In-Class Exercise
SK implementation has high TPR than TNR. test for the hypothesis: Do multiple runs and see the results are on average.
SK implementation has high TPR than TNR. test for the hypothesis: Do multiple runs and see the results are on average.

-- Oct 11 In-Class Exercise
Hypothesis: The accuracy of the SVM will improve as the amount of training data increases
Testing: We can test this by increasing the amount of training data in increments (e.g. 1000, 5000, 10000...) and comparing the accuracy after training
(Edited: 2017-10-11)
'''Hypothesis:''' The accuracy of the SVM will improve as the amount of training data increases '''Testing:''' We can test this by increasing the amount of training data in increments (e.g. 1000, 5000, 10000...) and comparing the accuracy after training

-- Oct 11 In-Class Exercise
Hypothesis 1: Increasing the size of the training data improves the accuracy. Description 1: Model can be trained and tested on different sizes of data and accuracy can be measured.
(Edited: 2017-10-11)
Hypothesis 1: Increasing the size of the training data improves the accuracy. Description 1: Model can be trained and tested on different sizes of data and accuracy can be measured.

-- Oct 11 In-Class Exercise
Hypothesis: It will take more number of training examples to learn complex symbols as compared to easier ones
Testing: Test this by using SVM as a bi-classifier and building model for each symbol and calculate accuracy for each case. Easier symbols will achieve high accuracy by using less number of training examples.
Hypothesis: It will take more number of training examples to learn complex symbols as compared to easier ones Testing: Test this by using SVM as a bi-classifier and building model for each symbol and calculate accuracy for each case. Easier symbols will achieve high accuracy by using less number of training examples.

-- Oct 11 In-Class Exercise
exp-1: Hyp-1: increasing the training data increase the accuracy of the model Hyp-2: if the ratio of +ve to -ve training data is more it adversely affect the performance of the model
exp-2: Hyp-1 the performance of model does not change when we are trying to identify any of the input zener images
exp-4: Hyp-1 :decrease in epsilon value is proportional to the increase in train accuracy but leads to decrease in test accuracy
(Edited: 2017-10-11)
exp-1: Hyp-1: increasing the training data increase the accuracy of the model Hyp-2: if the ratio of +ve to -ve training data is more it adversely affect the performance of the model exp-2: Hyp-1 the performance of model does not change when we are trying to identify any of the input zener images exp-4: Hyp-1 :decrease in epsilon value is proportional to the increase in train accuracy but leads to decrease in test accuracy

-- Oct 11 In-Class Exercise
Hypothesis: An SVM converges to an accurate value when the number of training samples increases
The hypothesis can be tested by training the model with a larger data set each time and comparing the accuracy
(Edited: 2017-10-11)
Hypothesis: An SVM converges to an accurate value when the number of training samples increases The hypothesis can be tested by training the model with a larger data set each time and comparing the accuracy

-- Oct 11 In-Class Exercise
Hypothesis 1 - SVM converges to a more accurate model when the amount of training data increases.
Testing - This hypothesis can be tested by increasing the amount of training data and comparing the accuracy.
Hypothesis 2 - Complex shapes require more training data as compared to simple shapes to get an accurate model.
Testing - This hypothesis can be tested by comparing the accuracy and no. of training samples used for different shapes.
(Edited: 2017-10-11)
Hypothesis 1 - SVM converges to a more accurate model when the amount of training data increases. Testing - This hypothesis can be tested by increasing the amount of training data and comparing the accuracy. Hypothesis 2 - Complex shapes require more training data as compared to simple shapes to get an accurate model. Testing - This hypothesis can be tested by comparing the accuracy and no. of training samples used for different shapes.

-- Oct 11 In-Class Exercise
Amer Rez
  • The best value for hyperparameter λ is x. I would run the algorithm a few times with different values for λ. for example we assign half of r/(r+ + r-) and then third of it. find which ones have better results (more accuracy or shorter training time)
(Edited: 2017-10-11)
Amer Rez * The best value for hyperparameter λ is x. I would run the algorithm a few times with different values for λ. for example we assign half of r/(r+ + r-) and then third of it. find which ones have better results (more accuracy or shorter training time)

-- Oct 11 In-Class Exercise
Name = Krishna Vojjila Hypothesis = increasing the training size will increase the accuracy of the model Testing = train and test the model with varying train sizes
Hypothesis 2 = the model will be dependent on the quality of train data and the class complexity and will take more time correspondingly Testing = train and test the model with simple classes(symbols) and then with complex classes
Name = Krishna Vojjila Hypothesis = increasing the training size will increase the accuracy of the model Testing = train and test the model with varying train sizes Hypothesis 2 = the model will be dependent on the quality of train data and the class complexity and will take more time correspondingly Testing = train and test the model with simple classes(symbols) and then with complex classes
[ Next ]
X