Instrumentation & Measurement Magazine 25-9 - 21

mainly related to olfactory learning, targeted experiments carried
out on the drosophila demonstrate involvement in more
complex tasks including attention; in addition, the models
built upon their basic identified structure can show even expectation
and delayed match-to-sample (Fig. 1). Leaving
further details on MBs to references [1], [4], the spiking activity
of the so-called kenyon cells (KCs), intrinsic neurons within
the main MB site called the calyx, can be represented as a lattice
of randomly connected neurons, via excitatory/inhibitory
synapses. KCs' neurons receive filtered olfactory inputs from
the antennal lobes (ALs), through a group of projection neurons,
encode such signals using spatio-temporal patterns of
neural activity and send their signals, through parallel axonal
fibers at the level of the so-called MB-lobes, to the output stage
for further conditioning (i.e., learning) activity. This influences
other brain structures, including the motor system, through
specific readout neural mappings.
The analysis of the MB structure was followed by a deep
modeling phase, after which neuro-inspired computational
models of a series of different behaviors were introduced, as
reported in Fig. 1. Some of them can be directly found in experiments
involving flies, such as classification, decision making,
distraction and attention. Furthermore, looking at the modeled
structure, other much more complex potential features
emerged. These were not (or not yet) discovered in fly experiments,
probably for the lack of a specific experimental setup,
but were detectable in larger insects, such as bees. This is the
case of sequence learning, which was successfully modeled by
adding accessory neural assemblies interacting with the main
MB structures and exploiting the reaction-diffusion effects
which can derive from the axo-axonal connection at the level
of the lobe system. In this paper, two examples are reviewed:
a state estimation task and a sequence and sub-sequence task.
MBs' Computational Model
With the aim to design and
reproduce the MB neural
dynamics, even at a scaleddown
size, a large degree
of similarity is found with
the reservoir computing
(RC) paradigm [5]. This is
based on recurrent neural
networks (RNNs)
showing hidden random
connections, with the characteristic
that only the
output layers are subject to
learning. RCs are adopted
to develop data-driven
models, especially when
nonlinear dynamic behaviors
are involved.
Two main classes of networks
were developed in
the literature adopting the
December 2022
RC paradigm: echo-state networks (ESN) and liquid state
machines (LSMs). These two solutions follow the same philosophy
in which the main difference relies on the nature of
the neuron model: this consists of a nonlinear static function
(e.g., hyperbolic tangent) in the case of ESNs and a more
biologically-inspired spiking dynamic system (e.g., leaky integrate-and-fire
neuron) for the LSMs.
An LSM-based (basic and augmented) solution will be reviewed
in this paper as a useful paradigm for implementing
spiking neural controllers for bio-inspired adaptive perception
on interacting artefacts. We can consider an LSM as a large
network of spiking neurons designed following the RC paradigm,
where different layers are considered: an encoding
layer to acquire input signals; a hidden recurrent layer where a
spatio-temporal activity emerges through nonlinear transformations
of the input data; and, finally, a decoding layer (i.e., a
read-out map) where the learning process is applied. By properly
setting the network parameters to guarantee the presence
of a fading memory and input separability, it has been demonstrated
that LSM is a universal function approximator.
The output layer can be trained using different learning
methods, including least mean square algorithms, FORCE
learning rule, genetic algorithms, and others [6].
The basic equations describing the LSM architecture,
shown in Fig. 2, are:

xt f xt u t
yt g xt
  


    

ut and the current internal state 
(1)
(2)
where f represents the nonlinear flow of the membrane potentials

xt of the reservoir layer consisting of interconnected
dynamical spiking systems. Its evolution depends on the input


xt . The output signals
yt are obtained through a memory-less function g that, in
Fig. 2. Scheme of a liquid state machine (LSM) where the different neural connections are identified.
IEEE Instrumentation & Measurement Magazine
21

Instrumentation & Measurement Magazine 25-9

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 25-9

Instrumentation & Measurement Magazine 25-9 - Cover1
Instrumentation & Measurement Magazine 25-9 - Cover2
Instrumentation & Measurement Magazine 25-9 - 1
Instrumentation & Measurement Magazine 25-9 - 2
Instrumentation & Measurement Magazine 25-9 - 3
Instrumentation & Measurement Magazine 25-9 - 4
Instrumentation & Measurement Magazine 25-9 - 5
Instrumentation & Measurement Magazine 25-9 - 6
Instrumentation & Measurement Magazine 25-9 - 7
Instrumentation & Measurement Magazine 25-9 - 8
Instrumentation & Measurement Magazine 25-9 - 9
Instrumentation & Measurement Magazine 25-9 - 10
Instrumentation & Measurement Magazine 25-9 - 11
Instrumentation & Measurement Magazine 25-9 - 12
Instrumentation & Measurement Magazine 25-9 - 13
Instrumentation & Measurement Magazine 25-9 - 14
Instrumentation & Measurement Magazine 25-9 - 15
Instrumentation & Measurement Magazine 25-9 - 16
Instrumentation & Measurement Magazine 25-9 - 17
Instrumentation & Measurement Magazine 25-9 - 18
Instrumentation & Measurement Magazine 25-9 - 19
Instrumentation & Measurement Magazine 25-9 - 20
Instrumentation & Measurement Magazine 25-9 - 21
Instrumentation & Measurement Magazine 25-9 - 22
Instrumentation & Measurement Magazine 25-9 - 23
Instrumentation & Measurement Magazine 25-9 - 24
Instrumentation & Measurement Magazine 25-9 - 25
Instrumentation & Measurement Magazine 25-9 - 26
Instrumentation & Measurement Magazine 25-9 - 27
Instrumentation & Measurement Magazine 25-9 - 28
Instrumentation & Measurement Magazine 25-9 - 29
Instrumentation & Measurement Magazine 25-9 - 30
Instrumentation & Measurement Magazine 25-9 - 31
Instrumentation & Measurement Magazine 25-9 - 32
Instrumentation & Measurement Magazine 25-9 - 33
Instrumentation & Measurement Magazine 25-9 - 34
Instrumentation & Measurement Magazine 25-9 - 35
Instrumentation & Measurement Magazine 25-9 - 36
Instrumentation & Measurement Magazine 25-9 - 37
Instrumentation & Measurement Magazine 25-9 - 38
Instrumentation & Measurement Magazine 25-9 - 39
Instrumentation & Measurement Magazine 25-9 - 40
Instrumentation & Measurement Magazine 25-9 - 41
Instrumentation & Measurement Magazine 25-9 - 42
Instrumentation & Measurement Magazine 25-9 - 43
Instrumentation & Measurement Magazine 25-9 - 44
Instrumentation & Measurement Magazine 25-9 - 45
Instrumentation & Measurement Magazine 25-9 - 46
Instrumentation & Measurement Magazine 25-9 - 47
Instrumentation & Measurement Magazine 25-9 - 48
Instrumentation & Measurement Magazine 25-9 - 49
Instrumentation & Measurement Magazine 25-9 - 50
Instrumentation & Measurement Magazine 25-9 - 51
Instrumentation & Measurement Magazine 25-9 - 52
Instrumentation & Measurement Magazine 25-9 - 53
Instrumentation & Measurement Magazine 25-9 - 54
Instrumentation & Measurement Magazine 25-9 - 55
Instrumentation & Measurement Magazine 25-9 - 56
Instrumentation & Measurement Magazine 25-9 - 57
Instrumentation & Measurement Magazine 25-9 - 58
Instrumentation & Measurement Magazine 25-9 - 59
Instrumentation & Measurement Magazine 25-9 - 60
Instrumentation & Measurement Magazine 25-9 - 61
Instrumentation & Measurement Magazine 25-9 - 62
Instrumentation & Measurement Magazine 25-9 - 63
Instrumentation & Measurement Magazine 25-9 - 64
Instrumentation & Measurement Magazine 25-9 - 65
Instrumentation & Measurement Magazine 25-9 - 66
Instrumentation & Measurement Magazine 25-9 - 67
Instrumentation & Measurement Magazine 25-9 - 68
Instrumentation & Measurement Magazine 25-9 - 69
Instrumentation & Measurement Magazine 25-9 - Cover3
Instrumentation & Measurement Magazine 25-9 - Cover4
https://www.nxtbook.com/allen/iamm/26-6
https://www.nxtbook.com/allen/iamm/26-5
https://www.nxtbook.com/allen/iamm/26-4
https://www.nxtbook.com/allen/iamm/26-3
https://www.nxtbook.com/allen/iamm/26-2
https://www.nxtbook.com/allen/iamm/26-1
https://www.nxtbook.com/allen/iamm/25-9
https://www.nxtbook.com/allen/iamm/25-8
https://www.nxtbook.com/allen/iamm/25-7
https://www.nxtbook.com/allen/iamm/25-6
https://www.nxtbook.com/allen/iamm/25-5
https://www.nxtbook.com/allen/iamm/25-4
https://www.nxtbook.com/allen/iamm/25-3
https://www.nxtbook.com/allen/iamm/instrumentation-measurement-magazine-25-2
https://www.nxtbook.com/allen/iamm/25-1
https://www.nxtbook.com/allen/iamm/24-9
https://www.nxtbook.com/allen/iamm/24-7
https://www.nxtbook.com/allen/iamm/24-8
https://www.nxtbook.com/allen/iamm/24-6
https://www.nxtbook.com/allen/iamm/24-5
https://www.nxtbook.com/allen/iamm/24-4
https://www.nxtbook.com/allen/iamm/24-3
https://www.nxtbook.com/allen/iamm/24-2
https://www.nxtbook.com/allen/iamm/24-1
https://www.nxtbook.com/allen/iamm/23-9
https://www.nxtbook.com/allen/iamm/23-8
https://www.nxtbook.com/allen/iamm/23-6
https://www.nxtbook.com/allen/iamm/23-5
https://www.nxtbook.com/allen/iamm/23-2
https://www.nxtbook.com/allen/iamm/23-3
https://www.nxtbook.com/allen/iamm/23-4
https://www.nxtbookmedia.com