In this paper, we propose EdgeHD, a hierarchy-aware learning solution that performs online training and inference in a highly distributed, cost-effective way. We use brain-inspired hyperdimensional (HD) computing as the key enabler. HD computing performs the computation tasks on a high-dimensional space to emulate functionalities of the human memory, such as inter-data relationship reasoning and information aggregation. EdgeHD exploits HD computing to effectively learn the classification models on individual devices and combine the models through the hierarchical IoT nodes without high communication costs. We also propose a hardware design that accelerates EdgeHD on low-power FPGA platforms. We evaluated EdgeHD for a wide range of real-world classification applications. The evaluation shows that EdgeHD provides highly efficient computation with reduced communication. For example, EdgeHD achieves on average$3.4\times$and$11.7\times (1.9\times$and$7.8\times$) speedup and energy efficiency improvement during the training (inference) as compared to the centralized learning approach. It reduces the communication costs by 85% for the training and 78% for the inference.

A critical step in how malaria parasites invade red blood cells (RBCs) is the wrapping of the membrane around the egg-shaped merozoites. Recent experiments have revealed that RBCs can be protected from malaria invasion by high membrane tension. While cellular and biochemical aspects of parasite actomyosin motor forces during the malaria invasion have been well studied, the important role of the biophysical forces induced by the RBC membrane-cytoskeleton composite has not yet been fully understood. In this study, we use a theoretical model for lipid bilayer mechanics, cytoskeleton deformation, and membrane-merozoite interactions to systematically investigate the influence of effective RBC membrane tension, which includes contributions from the lipid bilayer tension, spontaneous tension, interfacial tension, and the resistance of cytoskeleton against shear deformation on the progression of membrane wrapping during the process of malaria invasion. Our model reveals that this effective membrane tension creates a wrapping energy barrier for a complete merozoite entry. We calculate the tension threshold required to impede the malaria invasion. We find that the tension threshold is a nonmonotonic function of spontaneous tension and undergoes a sharp transition from large to small values as the magnitude of interfacial tension increases. We also predict that the physical properties of the RBC cytoskeleton layer-particularly the resting length of the cytoskeleton-play key roles in specifying the degree of the membrane wrapping. We also found that the shear energy of cytoskeleton deformation diverges at the full wrapping state, suggesting the local disassembly of the cytoskeleton is required to complete the merozoite entry. Additionally, using our theoretical framework, we predict the landscape of myosin-mediated forces and the physical properties of the RBC membrane in regulating successful malaria invasion. Our findings on the crucial role of RBC membrane tension in inhibiting malaria invasion can have implications for developing novel antimalarial therapeutic or vaccine-based strategies.

In the Internet of Things (IoT) domain, many applications are running machine learning algorithms to assimilate the data collected in the swarm of devices. Sending all data to the powerful computing environment, e.g., cloud, poses significant efficiency and scalability issues. A promising way is to distribute the learning tasks onto the IoT hierarchy, often referred to edge computing; however, the existing sophisticated algorithms such as deep learning are often overcomplex to run on less-powerful and unreliable embedded IoT devices. Hyperdimensional Computing (HDC) is a brain-inspired learning approach for efficient and robust learning on today's embedded devices. Encoding, or transforming the input data into high-dimensional representation, is the key first step of HDC before performing a learning task. All existing HDC approaches use a static encoder; thus, they still require very high dimensionality, resulting in significant efficiency loss for the edge devices with limited resources. In this paper, we have developed NeuralHD, a new HDC approach with a dynamic encoder for adaptive learning. Inspired by human neural regeneration study in neuroscience, NeuralHD identifies insignificant dimensions and regenerates those dimensions to enhance the learning capability and robustness. We also present a scalable learning framework to distribute NeuralHD computation over edge devices in IoT systems. Our solution enables edge devices capable of real-time learning from both labeled and unlabeled data. Our evaluation on a wide range of practical classification tasks shows that NeuralHD provides 5.7X and 6.1X (12.3X and 14.1X) faster and more energy-efficient training compared to the HD-based algorithms (DNNs) running on the same platform. NeuralHD also provides 4.2X and 11.6X higher robustness to noise in the unreliable network and hardware of IoT environments as compared to DNNs.

The biconcave disk shape of the mammalian red blood cell (RBC) is unique to the RBC and is vital for its circulatory function. Due to the absence of a transcellular cytoskeleton, RBC shape is determined by the membrane skeleton, a network of actin filaments cross-linked by spectrin and attached to membrane proteins. While the physical properties of a uniformly distributed actin network interacting with the lipid bilayer membrane have been assumed to control RBC shape, recent experiments reveal that RBC biconcave shape also depends on the contractile activity of nonmuscle myosin IIA (NMIIA) motor proteins. Here, we use the classical Helfrich-Canham model for the RBC membrane to test the role of heterogeneous force distributions along the membrane and mimic the contractile activity of sparsely distributed NMIIA filaments. By incorporating this additional contribution to the Helfrich-Canham energy, we find that the RBC biconcave shape depends on the ratio of forces per unit volume in the dimple and rim regions of the RBC. Experimental measurements of NMIIA densities at the dimple and rim validate our prediction that (a) membrane forces must be non-uniform along the RBC membrane and (b) the force density must be larger in the dimple than the rim to produce the observed membrane curvatures. Furthermore, we predict that RBC membrane tension and the orientation of the applied forces play important roles in regulating this force-shape landscape. Our findings of heterogeneous force distributions on the plasma membrane for RBC shape maintenance may also have implications for shape maintenance in different cell types.