How to do incremental training using 2 batches of dataset, where the tool is anything except neural network

1 view (last 30 days)
I am a novice in this field. I have two training datasets. 1st I want to train the machine with Dataset X (from fisherirish) which has 150 entries, After that I want to train the already trained model with a new dataset T2 , which has also 50 entries. Using FITCNB what i did was
X=meas
Y=specieserrorcopy
Mdl = fitcnb(X,Y)
label = predict(Mdl,X)
L = loss(Mdl,X,Y)
accERR=1-L
Please help me how to use this Mdl to train again using T2.

Accepted Answer

Greg Heath
Greg Heath on 17 Jan 2018
The new training will override the old.
Therefore you have to either
a. Add the old training set to the new one
or
b. Add representative examples of the new dataset to the old one.
Then retrain to obtain new coefficients.
Decades ago, the Army asked me to design a set of classifiers that would identify the different products resulting from a missile deployment (e.g., missile, decoys, & deployment hardware). New test flight data was delivered periodically.
I found the best approach was to use a radial basis function classifier. When new data was run through the classifier and misclassified, new clusters were added to the data base. Then new and old coefficients were updated to minimize error rates.
Hope this helps.
Thank you for formally accepting my answer
Greg
  2 Comments
debasmita bhoumik
debasmita bhoumik on 20 Jan 2018
Thank you sir. Please clarify one more thing. In my question I mentioned "except neural network". Now I understand there is no learning technique for incremental training except neural network. But now my confusion- IS IT POSSIBLE IN NEURAL NETWORK? Because I have tried to implement the above mentioned thing using Matlab App for neural network - train using one dataset and then another and got ans. Is it possible?
Greg Heath
Greg Heath on 21 Jan 2018
Ideally, you should know the SS (i.e., summary statistics, e.g, means, standard deviations and cross correlations) of all original and new training, validation and testing data.
For valid results the SS of all should be similar.
Otherwise, satisfactory performance cannot be guaranteed.
Hope this helps.
Greg

Sign in to comment.

More Answers (1)

debasmita bhoumik
debasmita bhoumik on 17 Jan 2018
Thank you sir. Please clarify one more thing. In my question I mentioned "except neural network". Now I understand there is no learning technique for incremental training except neural network. But now my confusion- IS IT POSSIBLE IN NEURAL NETWORK? Because I have tried to implement the above mentioned thing using Matlab App for neural network - train using one dataset and then another and got ans. Is it possible?

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!