Rewrite logistic regression model to use DataLoader interface to load data. about deeplearningresearch HOT 4 CLOSED

Workflow of new logistic regression:

1. Use Dataloader class to load data set.
2. Get required meta data from data set, eg. dimensionality
3. Create training model
4. Call trainModel(trainingSet, validationSet) to train the model
5. Call testModel(testSet) to test the model

from deeplearningresearch.

1. Use Dataloader class to load data set.
2. Get required meta data from data set, eg. dimensionality
3. Create training model
4. Call trainModel(trainingSet, validationSet) to train the model

These are done for logistic regression model. Next step:

1. refactor train model to use scan function.
2. encapsulate train method in another class, eg. earlystopping, pre-training
3. encapsulate testModel in a new function

from deeplearningresearch.

1. scan function seems not intent to replace for loop. I will look at more example to see if scan is more general used operation or just a new function. This requirement cancelled for now.
2. EarlyStopTrainer is uploaded. Pre-training will defer until I read the related paper.
3. Done. one method called testModel in Logistic Regression class. This should be a common interface for all models

from deeplearningresearch.

close

from deeplearningresearch.