This project has the goal of studying and proposing new algorithms concerning the output kernel in the Multi-task setting.
- Algorithms -
a) From the paper "Learning Output Kernels with Block Coordinate Descent", Algorithm 1 in "dinuzzo.py"
b) From the paper "Efficient Output Kernel Learning for Multiple Task", Algorithm 1 in "fast_MTL_SDCA-py"
- Datasets -
I) CIFAR-10 with features extracted by using Inception V2 from TensorFlow II) CIFAR-100 with features extracted by using Inception V2 from TensorFlow