From Event: SPIE Organic Photonics + Electronics, 2019
Unlike the binary logic of conventional circuitry, the brain relies on the adjustment of continuous synaptic weights to integrate complex stimuli, perform computations, and instruct an appropriate response.[1] Recently, organic electrochemical transistors (OECTs) have been established as an interesting analog synaptic mimic capable of exhibiting several memory functionalities.[2] The existing technology, however, still relies on static, pre-fabricated circuitry that cannot rewire itself in response to a stimulus.
Here, we report an evolvable OECT that is formed in situ by electropolymerizing a self-doped conjugated monomer as the transistor channel. Fabrication by means of electropolymerization allows for a stimulus-driven formation of new electronic synapses, which, in the biological sphere, is a major contributor to neuroplasticity. Lasting changes in channel properties can be achieved by either growing additional channel material to enhance conductance or by over-oxidizing the channel to reduce conductance, which is analogous to long-term potentiation and depression, respectively. Transient changes in channel conductance, analogous to short-term potentiation and depression, are attained by inducing non-equilibrium doping states within the transistor channel.
We incorporate the synaptic transistor into an evolvable learning circuit that acts as an electronic mimic of classical conditioning by linking an irrelevant input to a response by presenting it simultaneously with a relevant input. The simplicity of this circuit, which consists of a transistor in series with a resistor, is not attainable with existing technology. The reduction in complexity and footprint provided by evolvable hardware has potential to bring about a paradigm shift in the field of machine learning.
[1] J. I. Gold, M. N. Shadlen, Trends Cogn. Sci. 2001, 5, 10.
[2] a. Y. van de Burgt, E. Lubberman, E. J. Fuller, S. T. Keene, G. C. Faria, S. Agarwal, M. J. Marinella, A. A. Talin, A. Salleo, Nature Mater. 2017, 16, 414; b. P. Gkoupidenis, D. A. Koutsouras, G. G. Malliaras, Nature Comm. 2017, 8.
© (2019) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Jennifer Gerasimov, "An evolvable transistor as a basis for neuromorphic learning circuits (Conference Presentation)," Proc. SPIE 11096, Organic and Hybrid Sensors and Bioelectronics XII, 110960L (Presented at SPIE Organic Photonics + Electronics: August 12, 2019; Published: 10 September 2019); https://doi.org/10.1117/12.2524890.6083883348001.