Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors
Memristors have attracted interest as neuromorphic computation elements because they show promise in enabling efficient hardware implementations of artificial neurons and synapses. We performed measurements on interface-type memristors to validate their use in neuromorphic hardware. Specifically, we...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2021-02-01
|
| Series: | Frontiers in Neuroscience |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fnins.2020.627276/full |
| _version_ | 1819296906201792512 |
|---|---|
| author | Thomas F. Tiotto Thomas F. Tiotto Anouk S. Goossens Anouk S. Goossens Jelmer P. Borst Jelmer P. Borst Tamalika Banerjee Tamalika Banerjee Niels A. Taatgen Niels A. Taatgen |
| author_facet | Thomas F. Tiotto Thomas F. Tiotto Anouk S. Goossens Anouk S. Goossens Jelmer P. Borst Jelmer P. Borst Tamalika Banerjee Tamalika Banerjee Niels A. Taatgen Niels A. Taatgen |
| author_sort | Thomas F. Tiotto |
| collection | DOAJ |
| description | Memristors have attracted interest as neuromorphic computation elements because they show promise in enabling efficient hardware implementations of artificial neurons and synapses. We performed measurements on interface-type memristors to validate their use in neuromorphic hardware. Specifically, we utilized Nb-doped SrTiO3 memristors as synapses in a simulated neural network by arranging them into differential synaptic pairs, with the weight of the connection given by the difference in normalized conductance values between the two paired memristors. This network learned to represent functions through a training process based on a novel supervised learning algorithm, during which discrete voltage pulses were applied to one of the two memristors in each pair. To simulate the fact that both the initial state of the physical memristive devices and the impact of each voltage pulse are unknown we injected noise into the simulation. Nevertheless, discrete updates based on local knowledge were shown to result in robust learning performance. Using this class of memristive devices as the synaptic weight element in a spiking neural network yields, to our knowledge, one of the first models of this kind, capable of learning to be a universal function approximator, and strongly suggests the suitability of these memristors for usage in future computing platforms. |
| first_indexed | 2024-12-24T05:05:33Z |
| format | Article |
| id | doaj.art-cd3a3ff3539f44ebad06e414f90ccf14 |
| institution | Directory Open Access Journal |
| issn | 1662-453X |
| language | English |
| last_indexed | 2024-12-24T05:05:33Z |
| publishDate | 2021-02-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neuroscience |
| spelling | doaj.art-cd3a3ff3539f44ebad06e414f90ccf142022-12-21T17:13:49ZengFrontiers Media S.A.Frontiers in Neuroscience1662-453X2021-02-011410.3389/fnins.2020.627276627276Learning to Approximate Functions Using Nb-Doped SrTiO3 MemristorsThomas F. Tiotto0Thomas F. Tiotto1Anouk S. Goossens2Anouk S. Goossens3Jelmer P. Borst4Jelmer P. Borst5Tamalika Banerjee6Tamalika Banerjee7Niels A. Taatgen8Niels A. Taatgen9Groningen Cognitive Systems and Materials Center, University of Groningen, Groningen, NetherlandsArtificial Intelligence, Bernoulli Institute, University of Groningen, Groningen, NetherlandsGroningen Cognitive Systems and Materials Center, University of Groningen, Groningen, NetherlandsZernike Institute for Advanced Materials, University of Groningen, Groningen, NetherlandsGroningen Cognitive Systems and Materials Center, University of Groningen, Groningen, NetherlandsArtificial Intelligence, Bernoulli Institute, University of Groningen, Groningen, NetherlandsGroningen Cognitive Systems and Materials Center, University of Groningen, Groningen, NetherlandsZernike Institute for Advanced Materials, University of Groningen, Groningen, NetherlandsGroningen Cognitive Systems and Materials Center, University of Groningen, Groningen, NetherlandsArtificial Intelligence, Bernoulli Institute, University of Groningen, Groningen, NetherlandsMemristors have attracted interest as neuromorphic computation elements because they show promise in enabling efficient hardware implementations of artificial neurons and synapses. We performed measurements on interface-type memristors to validate their use in neuromorphic hardware. Specifically, we utilized Nb-doped SrTiO3 memristors as synapses in a simulated neural network by arranging them into differential synaptic pairs, with the weight of the connection given by the difference in normalized conductance values between the two paired memristors. This network learned to represent functions through a training process based on a novel supervised learning algorithm, during which discrete voltage pulses were applied to one of the two memristors in each pair. To simulate the fact that both the initial state of the physical memristive devices and the impact of each voltage pulse are unknown we injected noise into the simulation. Nevertheless, discrete updates based on local knowledge were shown to result in robust learning performance. Using this class of memristive devices as the synaptic weight element in a spiking neural network yields, to our knowledge, one of the first models of this kind, capable of learning to be a universal function approximator, and strongly suggests the suitability of these memristors for usage in future computing platforms.https://www.frontiersin.org/articles/10.3389/fnins.2020.627276/fullneuromorphic computingsupervised learninginterface memristorNb-doped SrTiO3neural networksspiking neural network |
| spellingShingle | Thomas F. Tiotto Thomas F. Tiotto Anouk S. Goossens Anouk S. Goossens Jelmer P. Borst Jelmer P. Borst Tamalika Banerjee Tamalika Banerjee Niels A. Taatgen Niels A. Taatgen Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors Frontiers in Neuroscience neuromorphic computing supervised learning interface memristor Nb-doped SrTiO3 neural networks spiking neural network |
| title | Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors |
| title_full | Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors |
| title_fullStr | Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors |
| title_full_unstemmed | Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors |
| title_short | Learning to Approximate Functions Using Nb-Doped SrTiO3 Memristors |
| title_sort | learning to approximate functions using nb doped srtio3 memristors |
| topic | neuromorphic computing supervised learning interface memristor Nb-doped SrTiO3 neural networks spiking neural network |
| url | https://www.frontiersin.org/articles/10.3389/fnins.2020.627276/full |
| work_keys_str_mv | AT thomasftiotto learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT thomasftiotto learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT anouksgoossens learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT anouksgoossens learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT jelmerpborst learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT jelmerpborst learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT tamalikabanerjee learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT tamalikabanerjee learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT nielsataatgen learningtoapproximatefunctionsusingnbdopedsrtio3memristors AT nielsataatgen learningtoapproximatefunctionsusingnbdopedsrtio3memristors |