M81: Short-Term Depression in VLSI Stochastic Synapse
P. Xu, T. Horiuchi, and P. Abshire ECE & Institute for Systems Research, University of Maryland, College Park, MD

· Emulate intrinsic stochastic behavior of biological synaptic transmission · Implement short/long term plasticity by modulating probability · Achieve spike level computation in probability domain
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.002 0.004 10 us 20 us 30 us 40 us 50 us

Vdd

Ibias Vc
Vicm Vi+

Vg+

M1 Vpre~

M2

Vg-

Vi- Vicm
Vtran Vdw

Vpre Vfw

Facilitation State Machine

VoM3

M5

Vo+ M4

Depression State Machine

p

Subtractive single release circuit model

1/r

0.006

0.008

0.01

Steady state probability  inverse input spiking rate r for different pulse widths

 v (t+ ) = v (t- ) - v, successful transmission at t  dv ( t )  = vmax - v (t ) between presynaptic spike  d dt   p ( t ) = f ( v ( t ))  1 pss  for av d r >> 1, p (t )  av (t ) r