Biomedical Engineering Reference
In-Depth Information
pdf at time index 45.092s
0.35
posterior density
desired velocity
velocity by seq. estimation (collapse)
velocity by seq. estimation (MLE)
velocity by adaptive filtering
0.3
0.25
0.2
0.15
0.1
0.05
0
-2.5
-2
-1.5
-1
-0.5
0
0.5
velocity
i
k
p
Δ
(
N
|
v
)
FIgURE 6.4:
at time 45.092 sec.
k
The velocity estimated by the sequential estimation with collapse denoted by the black circle is the
closest to the desired velocity (green star).
In summary, the Monte Carlo sequential estimation on point processes shows a good capabil-
ity to estimate the state from the discrete spiking events.
6.4 ENCodINg/dECodINg IN MoToR CoNTRol
It is interesting that in black box modeling, the motor BMI is posed as a decoding problem, that is, a
transformation from motor neurons to behavior. However, when we use the Bayesian sequential estima-
tion, decoding is insufficient to solve the modeling problem. The full dimension of the difficulty is still
partially hidden in the Kalman filter (how to update the state given the observations?), but it becomes
crystal clear when we develop the point process approach. To implement
decoding
, it is important to also
model how neurons
encode
movement. Therefore, one sees that generative models do in fact require more
information about the task and are therefore an opportunity to investigate further neural functionality.
One aspect that underlies the nervous system operation is the different time scales of the
neural spike trains (few milliseconds) versus behavior (motor system dynamics are at hundreds
