Future technology: machine learning using memristors networks
Add to your list(s)
Download to your calendar using vCal
If you have a question about this talk, please contact Christian Steinruecken.
I discuss the properties of general networks made of a class of memristors (resistors with memory) which are used for nonconventional computing. In fact, these components emulate the plasticity of neurons and can be fabricated in specialized laboratories. After having extensively introduced these components, we discuss a differential equation which describes the evolution of the internal memory of a generic circuit for ideal memristors.
This enables a formal treatment of the learning capability of these circuits. I then discuss the implications of such an equation for the use of memristors in machine learning, showing that in a certain limit of the parameter space the dynamics can be interpreted as a constrained gradient descent. I will also give a brief account of the formal connection to Statistical Mechanics at the end.
This talk is part of the Machine Learning @ CUED series.
This talk is included in these lists:
Note that exdirectory lists are not shown.
