Skip to content

Latest commit

 

History

History
7 lines (5 loc) · 958 Bytes

README.md

File metadata and controls

7 lines (5 loc) · 958 Bytes

MATML

Code used for my masters project on "Can machine learn?" Inlcudes code for stochastic gradient descdent (SGD), Model-Agnostic-Meta-Learning (MAML) and Model-Agnostic-Time-adjusted-Meta-Learning (MATML) algorithms. These algorithms are explained in detail in the report for my project also included in the repo. It also can carry out k-shot learning able to calculate alpha as explained in the report for both MAML and MATML.

Alongside these algorithms it also includes a class to generate neural networks and data from toy datasets including sin curve regression (one for generating sin curves with different amplitude and phase one for generating sin curves with a phase related to the time at which the sin curve is found) and binary gaussian classification (where the distributions shift over time).

It also includes many plotting functions alongside the ability to choose activation function, loss function, optimiser, data structure, etc...