HAS-RL: A Hierarchical Approximate Scheme Optimized With Reinforcement Learning for NoC-Based NN Accelerators

Network-on-Chip (NoC) is a scalable on-chip communication architecture for the NN accelerator, but with the increase in the number of nodes, the communication delay becomes higher. Applications such as machine learning have a certain resilience to noisy/erroneous transmitted data. Therefore, approximate communication becomes a promising solution...

Exploiting Symmetric Temporally Sparse BPTT for Efficient RNN Training

Recurrent Neural Networks (RNNs) are useful in temporal sequence tasks. However, training RNNs involves dense matrix multiplications which require hardware that can support a large number of arithmetic operations and memory accesses. Implementing online training of RNNs on the edge calls for optimized algorithms for an efficient deployment on...

Morphodynamic changes in the Yangtze Estuary under the impact of the Three Gorges Dam, estuarine engineering interventions and climate-induced sea level rise

The estuarine turbidity maximum (ETM) in the Yangtze Estuary Delta (YED) is muddy by definition and lacks bottom undulations. However, since 2013, a remarkable change has occurred in the YED. Recent images detected by a multibeam echosounder system, SeaBat 7125, for the first time have confirmed widespread regions of subaqueous dunes in the...

Revealing the Impact of Space-Charge Layers on the Li-Ion Transport in All-Solid-State Batteries

The influence of space-charge layers on the ionic charge transport over cathode-solid electrolyte interfaces in all-solid-state batteries remains unclear because of the difficulty to unravel it from other contributions to the ion transport over the interfaces. Here, we reveal the effect of the space-charge layers by systematically tuning the...