Computer Science

In-memory computing is a promising non-von Neumann approach for making energy-efficient deep learning inference hardware. Crossbar arrays of resistive memory devices can be used to encode the network weights and perform efficient analog matrix-vector multiplications without intermediate movements of data. However, due to device variability and noise, the network needs to be trained in a specific way so that transferring the digitally trained weights to the analog resistive memory devices will not result in significant loss of accuracy. Here, we introduce a methodology to train ResNet-type convolutional neural networks that results in no appreciable accuracy loss when transferring weights to in-memory computing hardware based on phase-change memory (PCM). We also propose a compensation technique that exploits the batch normalization parameters to improve the accuracy retention over time. We achieve a classification accuracy of 93.7% on the CIFAR-10 dataset and a top-1 accuracy on the ImageNet benchmark of 71.6% after mapping the trained weights to PCM. Our hardware results on CIFAR-10 with ResNet-32 demonstrate an accuracy above 93.5% retained over a one day period, where each of the 361,722 synaptic weights of the network is programmed on just two PCM devices organized in a differential configuration.

Starting with an experimentally observed networks of actin bundles, we model their network structure in terms of edges and nodes. We then compute and discuss the main electrical parameters, considering the bundles as electrical wires. A set of equations describing the network is solved with several initial conditions. Input voltages, that can be considered as information bits, are applied in a set of points and output voltages are computed in another set of positions. We consider both an idealized situation, where point-like electrodes can be inserted in any points of the bundles and a more realistic one, where electrodes lay on a surface and have typical dimensions available in the industry. We find that in both cases such a system can implement the main logical gates and a finite state machine.

The actin droplet machine is a computer model of a three-dimensional network of actin bundles developed in a droplet of a physiological solution, which implements mappings of sets of binary strings. The actin bundle network is conductive to travelling excitations, i.e. impulses. The machine is interfaced with an arbitrary selected set of k electrodes through which stimuli, binary strings of length k represented by impulses generated on the electrodes, are applied and responses are recorded. The responses are recorded in a form of impulses and then converted to binary strings. The machine's state is a binary string of length k: if there is an impulse recorded on the i th electrode, there is a `1' in the i-th position of the string, and `0' otherwise. We present a design of the machine and analyse its state transition graphs. We envisage that actin droplet machines could form an elementary processor of future massive parallel computers made from biopolymers.

Wearable devices are a fast-growing technology with impact on personal healthcare for both society and economy. Due to the widespread of sensors in pervasive and distributed networks, power consumption, processing speed, and system adaptation are vital in future smart wearable devices. The visioning and forecasting of how to bring computation to the edge in smart sensors have already begun, with an aspiration to provide adaptive extreme edge computing. Here, we provide a holistic view of hardware and theoretical solutions towards smart wearable devices that can provide guidance to research in this pervasive computing era. We propose various solutions for biologically plausible models for continual learning in neuromorphic computing technologies for wearable sensors. To envision this concept, we provide a systematic outline in which prospective low power and low latency scenarios of wearable sensors in neuromorphic platforms are expected. We successively describe vital potential landscapes of neuromorphic processors exploiting complementary metal-oxide semiconductors (CMOS) and emerging memory technologies (e.g. memristive devices). Furthermore, we evaluate the requirements for edge computing within wearable devices in terms of footprint, power consumption, latency, and data size. We additionally investigate the challenges beyond neuromorphic computing hardware, algorithms and devices that could impede enhancement of adaptive edge computing in smart wearable devices.

A Compressive Sensing (CS) approach is applied to utilize intrinsic computation capabilities of Spin-Orbit Torque Magnetic Random Access Memory (SOT-MRAM) devices for IoT applications wherein lifetime energy, device area, and manufacturing costs are highly-constrained while the sensing environment varies rapidly. In this manuscript, we propose the Adaptive Compressed-sampling via Multibit Crossbar Array (ACMCA) approach to intelligently generate the CS measurement matrix using a multibit SOT-MRAM crossbar array. SPICE circuit and MATLAB algorithm simulation results indicate that ACMCA reduces reconstruction Time-Averaged Normalized Mean Squared Error (TNMSE) by 5dB on average while providing up to 160 μ m 2 area reduction compared to a similar previous design presented in the literature while incurring a negligible increase in the energy consumption of generating the CS measurement matrix.

Photonic reservoir computing is an emergent technology toward beyond-Neumann computing. Although photonic reservoir computing provides superior performance in environments whose characteristics are coincident with the training datasets for the reservoir, the performance is significantly degraded if these characteristics deviate from the original knowledge used in the training phase. Here, we propose a scheme of adaptive model selection in photonic reservoir computing using reinforcement learning. In this scheme, a temporal waveform is generated by different dynamic source models that change over time. The system autonomously identifies the best source model for the task of time series prediction using photonic reservoir computing and reinforcement learning. We prepare two types of output weights for the source models, and the system adaptively selected the correct model using reinforcement learning, where the prediction errors are associated with rewards. We succeed in adaptive model selection when the source signal is temporally mixed, having originally been generated by two different dynamic system models, as well as when the signal is a mixture from the same model but with different parameter values. This study paves the way for autonomous behavior in photonic artificial intelligence and could lead to new applications in load forecasting and multi-objective control, where frequent environment changes are expected.

Using optical hardware for neuromorphic computing has become more and more popular recently due to its efficient high-speed data processing capabilities and low power consumption. However, there are still some remaining obstacles to realizing the vision of a completely optical neuromorphic computer. One of them is that, depending on the technology used, optical weighting elements may not share the same resolution as in the electrical domain. Moreover, noise and drift are important considerations as well. In this article, we investigate a new method for improving the performance of optical weighting, even in the presence of noise and in the case of very low resolution. Even with only 8 to 32 levels of resolution, the method can outperform the naive traditional low-resolution weighting by several orders of magnitude in terms of bit error rate and can deliver performance very close to full-resolution weighting elements, also in noisy environments.

The public access to noisy intermediate-scale quantum (NISQ) computers facilitated by IBM, Rigetti, D-Wave, etc., has propelled the development of quantum applications that may offer quantum supremacy in the future large-scale quantum computers. Parameterized quantum circuits (PQC) have emerged as a major driver for the development of quantum routines that potentially improve the circuit's resilience to the noise. PQC's have been applied in both generative (e.g. generative adversarial network) and discriminative (e.g. quantum classifier) tasks in the field of quantum machine learning. PQC's have been also considered to realize high fidelity quantum gates with the available imperfect native gates of a target quantum hardware. Parameters of a PQC are determined through an iterative training process for a target noisy quantum hardware. However, temporal variations in qubit quality metrics affect the performance of a PQC. Therefore, the circuit that is trained without considering temporal variations exhibits poor fidelity over time. In this paper, we present training methodologies for PQC in a completely classical environment that can improve the fidelity of the trained PQC on a target NISQ hardware by as much as 42.5%.

In this paper, we consider target detection in suspicious tissue via diffusive molecular communications (MCs). If a target is present, it continuously and with a constant rate secretes molecules of a specific type, so-called biomarkers, into the medium, which are symptomatic for the presence of the target. Detection of these biomarkers is challenging since due to the diffusion and degradation, the biomarkers are only detectable in the vicinity of the target. In addition, the exact location of the target within the tissue is not known. In this paper, we propose to distribute several reactive nanosensors (NSs) across the tissue such that at least some of them are expected to come in contact with biomarkers, which cause them to become activated. Upon activation, an NS releases a certain number of molecules of a secondary type into the medium to alert a fusion center (FC), where the final decision regarding the presence of the target is made. In particular, we consider a composite hypothesis testing framework where it is assumed that the location of the target and the biomarker secretion rate are unknown, whereas the locations of the NSs are known. We derive the uniformly most powerful (UMP) test for the detection at the NSs. For the final decision at the FC, we show that the UMP test does not exist. Hence, we derive a genie-aided detector as an upper bound on performance. We then propose two sub-optimal detectors and evaluate their performance via simulations

On the January 22nd 2019, Airbus launched a quantum computing challenge to solve a set of problems relevant for the aircraft life cycle (this https URL). The challenge consists of a set of 5 problems that ranges from design to deployment of aircraft. This work addresses the 5th problem. The formulation exploits an Integer programming framework with a linear objective function and the solution relies on the MemComputing paradigm. It is discussed how to use MemCPU TM software to solve efficiently the proposed problem and assess scaling properties, which turns out to be polynomial for meaningful solutions of the problem at hand. Also discussed are possible formulations of the problem utilizing non-linear objective functions, allowing for different optimization schemes implementable in modified MemCPU software, potentially useful for field operation purposes.