Eveline R. W. van Doremaele

Neural network training can be slow and energy-expensive due to the frequent transfer of weight data between digital memory and processing units. Neuromorphic systems can accelerate neural networks by performing multiply-accumulate operations in parallel using nonvolatile analog memory. However, executing the widely used backpropagation training algorithm in multilayer neural networks requires information—and therefore storage—of the partial derivatives of the weight values preventing suitable and scalable implementation in hardware. Here, we propose a hardware implementation of the backpropagation algorithm that progressively updates each layer using in situ stochastic gradient descent, avoiding this storage requirement. We experimentally demonstrate the in situ error calculation and the proposed progressive backpropagation method in a multilayer hardware-implemented neural network. We confirm identical learning characteristics and classification performance compared to conventional backpropagation in software. We show that our approach can be scaled to large and deep neural networks, enabling highly efficient training of advanced artificial intelligence computing systems.

Organic electrochemical transistors (OECTs) are of great interest in low‐power bioelectronics and neuromorphic computing, as they utilize organic mixed ionic‐electronic conductors (OMIECs) to transduce ionic signals into electrical signals. However, the poor environmental stability of OMIEC materials significantly restricts the practical application of OECTs. Therefore, the non‐fused planar naphthalenediimide (NDI)‐dialkoxybithiazole (2Tz) copolymers are fine‐tuned through varying ethylene glycol (EG) side chain lengths from tri(ethylene glycol) to hexa(ethylene glycol) (namely P‐XO, X = 3–6) to achieve OECTs with high‐stability and low threshold voltage. As a result, the NDI‐2Tz copolymers exhibit ambipolarity, rapid response (5. This work sheds light on the design of exceptional n‐type/ambipolar materials for OECTs. It demonstrates the potential of incorporating these ambipolar polymers into water‐operational integrated circuits for long‐term biosensing systems and energy‐efficient brain‐inspired computing.

Signal communication mechanisms within the human body rely on the transmission and modulation of action potentials. Replicating the interdependent functions of receptors, neurons and synapses with organic artificial neurons and biohybrid synapses is an essential first step towards merging neuromorphic circuits and biological systems, crucial for computing at the biological interface. However, most organic neuromorphic systems are based on simple circuits which exhibit limited adaptability to both external and internal biological cues, and are restricted to emulate only specific the functions of an individual neuron/synapse. Here, we present a modular neuromorphic system which combines organic spiking neurons and biohybrid synapses to replicate a neural pathway. The spiking neuron mimics the sensory coding function of afferent neurons from light stimuli, while the neuromodulatory activity of interneurons is emulated by neurotransmitters-mediated biohybrid synapses. Combining these functions, we create a modular connection between multiple neurons to establish a pre-processing retinal pathway primitive.

The fundamental mechanisms of signal communication within the human body rely on the spiking frequency of action potentials.^1,2 Through biological receptors and afferent neuronal cells, stimuli from the external world are encoded into a spiking pattern and transmitted to the central nervous systems where they are processed via interneurons. Replicating the interdependent functions of receptors, afferent neurons and interneurons with spiking circuits ¹, sensors³ and biohybrid synapses⁴ is an essential first step towards merging neuromorphic circuits and biological systems, crucial for computing at the biological interface. We present a novel adaptive spiking circuit that replicates afferent neurons sensory coding from external physical stimuli. We emulate the neuromodulatory activity of interneurons by associating the spiking circuit with biohybrid synapses demonstrating an interdependent chemical synaptic connection. To establish a full neuronal pathway, we combine these key biological functions, showing the signal transduction from light stimulus to spiking frequency and to dopamine-mediated plasticity: a retinal pathway primitive.

Organic mixed ionic–electronic conductors (OMIECs) are central to bioelectronic applications such as biosensors, health‐monitoring devices, and neural interfaces, and have facilitated efficient next‐generation brain‐inspired computing and biohybrid systems. Despite these examples, smart and adaptive circuits that can locally process and optimize biosignals have not yet been realized. Here, a tunable sensing circuit is shown that can locally modulate biologically relevant signals like electromyograms (EMGs) and electrocardiograms (ECGs), that is based on a complementary logic inverter combined with a neuromorphic memory element, and that is constructed from a single polymer mixed conductor. It is demonstrated that a small neuromorphic array based on this material effects high classification accuracy in heartbeat anomaly detection. This high‐performance material allows for straightforward monolithic integration, which reduces fabrication complexity while also achieving high on/off ratios with excellent ambient p‐ and n‐type stability in transistor performance. This material opens a route toward simple and straightforward fabrication and integration of more sophisticated adaptive circuits for future smart bioelectronics.

Organic electrochemical transistors (OECTs) have emerged as building blocks for low power circuits, biosensors, and neuromorphic computing. While p‐type polymer materials for OECTs are well developed, the choice of high‐performance n‐type polymers is limited, despite being essential for cation and metabolite biosensors, and crucial for constructing complementary circuits. N‐type conjugated polymers that have efficient ion‐to‐electron transduction are highly desired for electrochemical applications. In this contribution, three non‐fused, planar naphthalenediimide (NDI)‐dialkoxybithiazole (2Tz) copolymers, which systematically increase the amount of polar tri(ethylene glycol) (TEG) side chains: PNDI2OD‐2Tz (0 TEG), PNDIODTEG‐2Tz (1 TEG), PNDI2TEG‐2Tz (2 TEG), are reported. It is demonstrated that the OECT performance increases with the number of TEG side chains resulting from the progressively higher hydrophilicity and larger electron affinities. Benefiting from the high electron mobility, excellent ion conduction capability, efficient ion‐to‐electron transduction, and low‐lying lowest unoccupied molecular orbital energy level, the 2 TEG polymer achieves close to 10⁵ on‐off ratio, fast switching, 1000 stable operation cycles in aqueous electrolyte, and has a long shelf life. Moreover, the higher number TEG chain substituted polymer exhibits good conductance state retention over two orders of magnitudes in electrochemical resistive random‐access memory devices, highlighting its potential for neuromorphic computing.

We present an overview of the latest studies on organic neuromorphic and smart sensing devices and highlight the potential of these concepts to enhance the interaction efficiency between electronics and biological substances.

In this work, we report on the flow effects induced by a nanosecond pulsed helium cold atmospheric pressure plasma jet impinging on a glass substrate. Gas temperature measurements have been performed using optical emission spectroscopy including rotational spectra of OH(A-X) (0,0) and N ₂ (C-B) (0,0) and resonance broadening of a helium emission line at 667.8 nm. The measured increase in gas temperature was less than 15 K. We have measured the air concentration distribution in the jet effluent by means of Rayleigh scattering exploiting the large difference in Rayleigh cross-section of air and helium. The obtained results show that the plasma causes a broadening of the helium channel suggesting an enhanced gas velocity and mixing with the ambient air. The impact of the plasma on the jet effluent is polarity dependent and is larger in the case of positive applied voltage pulses. Using an estimation of the increased gas velocity required to obtain the observed air concentrations, we have shown that both gas heating and electrohydrodynamic (EHD) forces related to the propagation of an ionization front are not causing the observed phenomena. The surface charge which can exist for a much longer time than the duration of the applied voltage pulse is the primary cause of the EHD forces on the bulk charges which lead to significant different air concentration profiles in the jet effluent for plasma on and off.