predict resting-state using saved model

[zhangyu2ustc]

as one type of transfer learning, we will save the best model from ALL tasks, and predict the state transition in resting-state data

[zhangyu2ustc]

First, retrain the GCN model and save the best one:

''' from . import checkmat as checkmate best_ckpt_saver = checkmate.BestCheckpointSaver(save_dir=path,num_to_keep=3,maximize=True,saver=self.op_saver) best_ckpt_saver.handle(accuracy, sess, self.global_step) '''

Problem: it deletes some of the models because of conflicting of existing saver: ''' self.op_saver.save(sess, path, global_step=self.global_step) '''

[zhangyu2ustc]

the most common states from resting-state prediction are: motor, language and social

[zhangyu2ustc]

guide to restore any pre-trained model and use it for prediction, fine-tuning or further training: https://cv-tricks.com/tensorflow-tutorial/save-restore-tensorflow-models-quick-complete-tutorial/