![Page 1: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/1.jpg)
Neuromorphic computing with magnetic nano-oscillators
Julie Grollier1
Philippe Talatchian1, Miguel Romera1, Sumito Tsunegi2, Flavio Abreu Araujo1, Mathieu Riou 1, Jacob Torrejon 1, Vincent Cros1, Alice Mizrahi 1, Paolo
Bortolotti1, Juan Trastoy1, Kay Yakushiji2, Akio Fukushima2, Hitoshi Kubota2, Shinji Yuasa2, Maxence Ernoult1,3, Damir Vodenicarevic3, Tifenn Hirtzlin3,
Nicolas Locatelli3, Damien Querlioz3
1CNRS/Thales, France 2AIST, Japan 3C2N, France
bioSPINspired
![Page 2: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/2.jpg)
Neural networks run on unoptimized hardware
2
0011011
GPUs, TPUs, FPGAsneurons neurons
synapses
![Page 3: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/3.jpg)
Entangling memory and processing allows for fastand energy efficient computing
3
proces-sing
memory
Digital computer
synapse
neuron neuronmemory
processing
Brain
20 W in total !
processing
100 W/cm2
CPUs, GPUs, TPUs, FPGAs
![Page 4: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/4.jpg)
Can we build small neuromorphic chips that rundeep neural networks ?
4
Hundred millions of neurons and synapses in a 1 cm2 chip Each device smaller than 1 µm2
Nanoneurons
NanoneuronsNano-synapses
![Page 5: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/5.jpg)
5Moritz Helmstaedter lab, retina flight 2013
Future AI will be massively interconnected
Brain: 104 synapses/neurons
![Page 6: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/6.jpg)
Main trend : CMOS neurons + Memristive synapses
6
HP labsmemristor
CMOS
V1V2Vi…
G1
G2
Gi
I = S Gi Vi
10 000 synapses per neuron ?
Strukov and Williams, PNAS 106, 20155 (2009)
![Page 7: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/7.jpg)
Wireless deep learning through radio-frequency (RF) communications
7
![Page 8: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/8.jpg)
Magnetic nano-oscillators are non-linear nano-radios
8
CoFeB
FeB
MgO
spin torque
N. Locatelli, V. Cros and J. Grollier, Spin-torque building blocks, Nature Mat. 13, 11 (2014)
magnetic tunnel junction
compatible with CMOS
10-100 nm
Nanoscale, fast (GHz), non-linear and easily measurable
Same structure as magnetic memories
current0 10 20
-10
-5
0
5
10
Time (ns)
Voltage (
mV
)
)
0 2 4 6 8 10
0
5
10
15
Voltag
e a
mplit
ude (
mV
)
dc current (mA)
![Page 9: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/9.jpg)
Due to its stability and non-linearity, a single magnetic oscillator can emulate an assembly of neurons and perform neuromorphic computing
9J. Torrejon, M. Riou, F. Abreu Araujo et al, Nature 547, 428 (2017)
Spoken digit recognition through reservoir computing
![Page 10: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/10.jpg)
Magnetic nano-oscillators have a high tunability : theyare radio-receivers
10
AC signals
Enhanced sync ranges
0.0 0.5 1.0 1.5 2.0
3.4
3.6
3.8
4.0
fre
que
ncy (
GH
z)
dc current (mA)0 25 50 75 100 125
2.5
3.0
3.5
4.0
4.5
fre
que
ncy (
GH
z)
magnetic field (Oe)
A. Slavin and V. Tiberkevich, IEEE TM 45, 1875 (2009)
![Page 11: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/11.jpg)
The oscillators ability to mutually interact opens the path to RF on-chip communication between neuronlayers
11
…
Same frequency: strong synapse
Different frequencies: weak synapse
i j
![Page 12: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/12.jpg)
Vowels classification with spin-torque oscillator neural network
Inputs: vowels
MagneticNano-Oscillators
Outputs: synchronized states
fA fB
M. Romera, P. Talatchian et al,Nature 563, 230 (2018)
![Page 13: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/13.jpg)
13320 340 360 380
10-4
10-2
100
Spectr
al pow
er
density (
µW
/MH
z)
Frequency (MHz)
I1 I2 I3 I4
Response of the neural network without inputs
![Page 14: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/14.jpg)
320 340 360 380
10-4
10-2
100
Spectr
al pow
er
density (
µW
/MH
z)
Frequency (MHz)
fA fB
I1 I2 I3 I4
The inputs modify the oscillator responses:oscillator 4 is sync to source B
fA fB
![Page 15: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/15.jpg)
15
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
We summarize all thesemeasurements in a map
where the differentsynchronization states have different colors
4B
1A
3A
4B
![Page 16: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/16.jpg)
16
Microwave sources~ 300 MHz
MagneticNano-Oscillators~ 300 MHz
Outputs: synchronized states
fA fB
Hillebrands Michingandatabase
Input vowels~ 1 kHz 12 formant frequencies + 1 bias
![Page 17: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/17.jpg)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
For classification, all the points
corresponding to one vowel should fall in a single synchronization
region
![Page 18: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/18.jpg)
We train the network by tuning the currents throughthe oscillators according to an online learning rule
18
I1 I2 I3 I4
Oscillators
1-100
-80
-60
-40
-20
350 370
am
plit
ude (
dB
m)
frequency (MHz)330
2 3 4
fA fB
ex: « iy » 4B
Experimental set-up
Computer(Learning algorithm)
![Page 19: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/19.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 1 step:
0
20
40
60
80
100
4
6
8
10
0 20 40 60 80
340
350
360
370
Number of training steps
![Page 20: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/20.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 3 steps:
0
20
40
60
80
100
4
6
8
10
0 20 40 60 80
340
350
360
370
Number of training steps
![Page 21: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/21.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 5 steps:
0
20
40
60
80
100
4
6
8
10
0 20 40 60 80
340
350
360
370
Number of training steps
![Page 22: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/22.jpg)
4
6
8
10
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 7 steps:
0
20
40
60
80
100
![Page 23: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/23.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 9 steps:
0
20
40
60
80
100
4
6
8
10
![Page 24: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/24.jpg)
4
6
8
10
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 12 steps:
0
20
40
60
80
100
![Page 25: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/25.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 15 steps:
0
20
40
60
80
100
4
6
8
10
![Page 26: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/26.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 18 steps:
0
20
40
60
80
100
4
6
8
10
![Page 27: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/27.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 21 steps:
0
20
40
60
80
100
4
6
8
10
![Page 28: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/28.jpg)
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 29 steps:
0
20
40
60
80
100
4
6
8
10
![Page 29: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/29.jpg)
4
6
8
10
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 35 steps:
0
20
40
60
80
100
![Page 30: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/30.jpg)
4
6
8
10
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
After 40 steps:
0
20
40
60
80
100
recognition rate (%)
![Page 31: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/31.jpg)
4
6
8
10
oscillator frequencies (MHz)
dc currents in oscillators(mA)
0 20 40 60 80
340
350
360
370
Number of training steps
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
After 44 steps:
0
20
40
60
80
100
recognition rate (%)
![Page 32: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/32.jpg)
4
6
8
10
0 20 40 60 80
340
350
360
370
Number of training steps
oscillator frequencies (MHz)
dc currents in oscillators(mA)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
recognition rate (%)After 48 steps:
0
20
40
60
80
100
![Page 33: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/33.jpg)
0
20
40
60
80
100
4
6
8
10
0 20 40 60 80
340
350
360
370
Number of training steps
oscillator frequencies (MHz)
dc currents in oscillators(mA)
fB(MHz)
380
370
360
350
340
330
fA (MHz)330 340 350 360 370 380
After 86 steps: recognition rate (%)
85 %
![Page 34: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/34.jpg)
34
What’s next ?
![Page 35: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/35.jpg)
Deep neural networks that communicate throughmicrowaves
35
…
RF Nano-neurons
RF Nano-neuronsRF Nano-
Synapses
![Page 36: Neuromorphic computing with magnetic nano-oscillators julie.pdfNeuromorphic computing with magnetic nano-oscillators Julie Grollier1 Philippe Talatchian 1, Miguel Romera , Sumito Tsunegi2,](https://reader033.vdocuments.net/reader033/viewer/2022051807/6008400086c97618364275d2/html5/thumbnails/36.jpg)
36