Thank you for sharing the paper. I had two questions. I am interested in knowing what are the different mechanisms that would allow integration of 'transfer learning' with XdG and synaptic stabilization. How would this combination alleviate 'catastrophic forgetting'? Would the accuracy be greater than the one achieved with only XdG and synaptic stabilization?
My second question is whether you tested if the gradual loss in accuracy (90.7%) with increasing sequentially trained tasks is monotonically decreasing or does it saturate to a stable equilibrium at some parameter value?
Thank you for sharing the paper. I had two questions. I am interested in knowing what are the different mechanisms that would allow integration of 'transfer learning' with XdG and synaptic stabilization. How would this combination alleviate 'catastrophic forgetting'? Would the accuracy be greater than the one achieved with only XdG and synaptic stabilization? My second question is whether you tested if the gradual loss in accuracy (90.7%) with increasing sequentially trained tasks is monotonically decreasing or does it saturate to a stable equilibrium at some parameter value?