A Rényi Intézet Deep Learning szemináriumának következő előadását Nyíri Tamás Bence (ELTE IK) tartja november 16-án szerdán 16:00-kor. Az előadás magyar nyelven, hibrid formában kerül megtartásra: mindenkit szeretettel várunk a Rényi Intézet Tondós termében, de az előadás az alábbi linken is követhető lesz:
In this talk we are trying to convince the audience that deeper models and larger datasets should not be the sole focus of machine learning, even though this strategy has done very well in recent times and that deep learning on small datasets will play a significant role moving forward.
We will first look at several competing definitions of 'small data', then give a motivation by citing real world scenarios where dealing with small data is inevitable.
After this, we take a short detour to look at research in human development and neurology to prove the biological capability of high quality learning from tiny amounts of data.
Given the feasibility of such a project, we aim to take a look at roadblocks preventing current deep learning methods from achieving such results.
Finally, we list the currently most prominent techniques that try to overcome such challenges and give examples on their usage in real world supervised learning tasks.
