RESUMO
Over the past few years, several applications have been extensively exploiting the advantages of deep learning, in particular when using convolutional neural networks (CNNs). The intrinsic flexibility of such models makes them widely adopted in a variety of practical applications, from medical to industrial. In this latter scenario, however, using consumer Personal Computer (PC) hardware is not always suitable for the potential harsh conditions of the working environment and the strict timing that industrial applications typically have. Therefore, the design of custom FPGA (Field Programmable Gate Array) solutions for network inference is gaining massive attention from researchers and companies as well. In this paper, we propose a family of network architectures composed of three kinds of custom layers working with integer arithmetic with a customizable precision (down to just two bits). Such layers are designed to be effectively trained on classical GPUs (Graphics Processing Units) and then synthesized to FPGA hardware for real-time inference. The idea is to provide a trainable quantization layer, called Requantizer, acting both as a non-linear activation for neurons and a value rescaler to match the desired bit precision. This way, the training is not only quantization-aware, but also capable of estimating the optimal scaling coefficients to accommodate both the non-linear nature of the activations and the constraints imposed by the limited precision. In the experimental section, we test the performance of this kind of model while working both on classical PC hardware and a case-study implementation of a signal peak detection device running on a real FPGA. We employ TensorFlow Lite for training and comparison, and use Xilinx FPGAs and Vivado for synthesis and implementation. The results show an accuracy of the quantized networks close to the floating point version, without the need for representative data for calibration as in other approaches, and performance that is better than dedicated peak detection algorithms. The FPGA implementation is able to run in real time at a rate of four gigapixels per second with moderate hardware resources, while achieving a sustained efficiency of 0.5 TOPS/W (tera operations per second per watt), in line with custom integrated hardware accelerators.
RESUMO
The phosphorylation and dephosphorylation of the NF-AT family of transcription factors play a key role in the activation of T lymphocytes and in the control of the immune response. The mechanistic aspects of NF-AT4 phosphorylation by protein kinase CK1 have been studied in this work with the aid of a series of 27 peptides, reproducing with suitable modifications the regions of NF-AT4 that have been reported to be phosphorylated by this protein kinase. The largest parent peptide, representing the three regions A, Z, and L spanning amino acids 173-218, is readily phosphorylated by CK1 at seryl residues belonging to the A2 segment, none of which fulfill the canonical consensus sequence for CK1. An acidic cluster of amino acids in the linker region between domains A and Z is essential for high-efficiency phosphorylation of the A2 domain, as shown by the increase in K(m) caused by a deletion of the linker region or a substitution of the acidic residues with glycines. Individual substitutions with alanine of each of the five serines in the A2 domain (S-177, S-180, S-181, S-184, and S-186) reduce the phosphorylation rate, the most detrimental effect being caused by Ser177 substitution which results in a 10-fold drop in V(max). On the contrary, the replacement of Ser177 with phosphoserine triggers a hierarchical effect with a dramatic improvement in phosphorylation efficiency, which no longer depends on the linker region for optimal efficiency. These data are consistent with a two-phase phosphorylation mechanism of NF-AT4 by CK1, initiated by the linker region which provides a functional docking site for CK1 and allows the unorthodox phosphorylation of Ser177; once achieved, this phosphoserine residue primes the phosphorylation of other downstream seryl residues, according to a hierarchical mechanism typically exploited by CK1.
