Is Learning Invisible to the Learner?

You can not see what you learn from learners. I remember all the things I learned when I was a child and see if I remember it. If you truly remember them, think about how you learned them. Have you noticed that your experience is with you today too? Or are you aware of what you are learning from the companies around you? Then those who better understand how they learned and what they learned can become better learners. If you understand better the way you learned and what you learned, you will become a better learner.

Please know your learning style. Someone learns best by seeing the teaching materials (visually learners), some learn the best (hearing learner) by listening to the teaching materials, and some will do best (sensory learner). In many cases, the more you feel involved, the more you learn, the more types you use to learn, the easier it is to remember. Organize your study time: Before you start plan ahead. Determine exactly what you want to achieve. Always keep more time than you think you need and prioritize your work to make sure you have enough time to deal with the most important things. Do not forget to do hard work when you are most cautious and fresh, and take a short break.

Meta-learning has meta-learners and learners. A meta-learner (or agent) trains a learner (or model) with a training set that contains many different tasks. At this stage of meta-learning, the model builds on previous experience from training and learns the common function of all tasks. Then whenever you need to learn a new task, fine-tune the model using the previous experience, using a small amount of new training data from the task. However, I do not want to start with randomly initializing parameters. This is because performance will not improve if only some updates are made for each task.

Conceptually, meta-learning is a learning process. Meta-learning algorithms usually accept task assignments. Each task is a learning problem and creates a fast learner, a learner that can be driven from a few examples. Meta-learning is a new research field in the field of deep learning. Last year, Berkeley Research Lab developed a new meta-learning algorithm called Model Agnostic Meta-Learning (MAML) that sets space standards. A couple days ago, OpenAI researchers published a paper describing a new meta learning algorithm called Reptilian.