Scalable Multi-FPGA HPC Architecture for Associative Memory System.

Associative memory is a cornerstone of cognitive intelligence within the human brain. The Bayesian confidence propagation neural network (BCPNN), a cortex-inspired model with high biological plausibility, has proven effective in emulating high-level cognitive functions like associative memory. However, the current approach using GPUs to simulate BCPNN-based associative memory tasks encounters challenges in latency and power efficiency as the model size scales. This work proposes a scalable multi-FPGA high performance computing (HPC) architecture designed for the associative memory system. The architecture integrates a set of hypercolumn unit (HCU) computing cores for intra-board online learning and inference, along with a spike-based synchronization scheme for inter-board communication among multiple FPGAs. Several design strategies, including population-based model mapping, packet-based spike synchronization, and cluster-based timing optimization, are presented to facilitate the multi-FPGA implementation. The architecture is implemented and validated on two Xilinx Alveo U50 FPGA cards, achieving a maximum model size of 200×10 and a peak working frequency of 220 MHz for the associative memory system. Both the memory-bounded spatial scalability and compute-bounded temporal scalability of the architecture are evaluated and optimized, achieving a maximum scale-latency ratio (SLR) of 268.82 for the two-FPGA implementation. Compared to a two-GPU counterpart, the two-FPGA approach demonstrates a maximum latency reduction of 51.72× and a power reduction exceeding 5.28× under the same network configuration. Compared with the state-of-the-art works, the two-FPGA implementation exhibits a high pattern storage capacity for the associative memory task.

A Memristor-Based Learning Engine for Synaptic Trace-Based Online Learning.

The memristor has been extensively used to facilitate the synaptic online learning of brain-inspired spiking neural networks (SNNs). However, the current memristor-based work can not support the widely used yet sophisticated trace-based learning rules, including the trace-based Spike-Timing-Dependent Plasticity (STDP) and the Bayesian Confidence Propagation Neural Network (BCPNN) learning rules. This paper proposes a learning engine to implement trace-based online learning, consisting of memristor-based blocks and analog computing blocks. The memristor is used to mimic the synaptic trace dynamics by exploiting the nonlinear physical property of the device. The analog computing blocks are used for the addition, multiplication, logarithmic and integral operations. By organizing these building blocks, a reconfigurable learning engine is architected and realized to simulate the STDP and BCPNN online learning rules, using memristors and 180 nm analog CMOS technology. The results show that the proposed learning engine can achieve energy consumption of 10.61 pJ and 51.49 pJ per synaptic update for the STDP and BCPNN learning rules, respectively, with a 147.03× and 93.61× reduction compared to the 180 nm ASIC counterparts, and also a 9.39× and 5.63× reduction compared to the 40 nm ASIC counterparts. Compared with the state-of-the-art work of Loihi and eBrainII, the learning engine can reduce the energy per synaptic update by 11.31× and 13.13× for trace-based STDP and BCPNN learning rules, respectively.

Mapping the BCPNN Learning Rule to a Memristor Model.

The Bayesian Confidence Propagation Neural Network (BCPNN) has been implemented in a way that allows mapping to neural and synaptic processes in the human cortexandhas been used extensively in detailed spiking models of cortical associative memory function and recently also for machine learning applications. In conventional digital implementations of BCPNN, the von Neumann bottleneck is a major challenge with synaptic storage and access to it as the dominant cost. The memristor is a non-volatile device ideal for artificial synapses that fuses computation and storage and thus fundamentally overcomes the von Neumann bottleneck. While the implementation of other neural networks like Spiking Neural Network (SNN) and even Convolutional Neural Network (CNN) on memristor has been studied, the implementation of BCPNN has not. In this paper, the BCPNN learning rule is mapped to a memristor model and implemented with a memristor-based architecture. The implementation of the BCPNN learning rule is a mixed-signal design with the main computation and storage happening in the analog domain. In particular, the nonlinear dopant drift phenomenon of the memristor is exploited to simulate the exponential decay of the synaptic state variables in the BCPNN learning rule. The consistency between the memristor-based solution and the BCPNN learning rule is simulated and verified in Matlab, with a correlation coefficient as high as 0.99. The analog circuit is designed and implemented in the SPICE simulation environment, demonstrating a good emulation effect for the BCPNN learning rule with a correlation coefficient as high as 0.98. This work focuses on demonstrating the feasibility of mapping the BCPNN learning rule to in-circuit computation in memristor. The feasibility of the memristor-based implementation is evaluated and validated in the paper, to pave the way for a more efficient BCPNN implementation, toward a real-time brain emulation engine.

Optimizing BCPNN Learning Rule for Memory Access.

Simulation of large scale biologically plausible spiking neural networks, e.g., Bayesian Confidence Propagation Neural Network (BCPNN), usually requires high-performance supercomputers with dedicated accelerators, such as GPUs, FPGAs, or even Application-Specific Integrated Circuits (ASICs). Almost all of these computers are based on the von Neumann architecture that separates storage and computation. In all these solutions, memory access is the dominant cost even for highly customized computation and memory architecture, such as ASICs. In this paper, we propose an optimization technique that can make the BCPNN simulation memory access friendly by avoiding a dual-access pattern. The BCPNN synaptic traces and weights are organized as matrices accessed both row-wise and column-wise. Accessing data stored in DRAM with a dual-access pattern is extremely expensive. A post-synaptic history buffer and an approximation function thus are introduced to eliminate the troublesome column update. The error analysis combining theoretical analysis and experiments suggests that the probability of introducing intolerable errors by such optimization can be bounded to a very small number, which makes it almost negligible. Derivation and validation of such a bound is the core contribution of this paper. Experiments on a GPU platform shows that compared to the previously reported baseline simulation strategy, the proposed optimization technique reduces the storage requirement by 33%, the global memory access demand by more than 27% and DRAM access rate by more than 5%; the latency of updating synaptic traces decreases by roughly 50%. Compared with the other similar optimization technique reported in the literature, our method clearly shows considerably better results. Although the BCPNN is used as the targeted neural network model, the proposed optimization method can be applied to other artificial neural network models based on a Hebbian learning rule.

Ribose-protonated DNA base excision repair: a combined theoretical and experimental study.

Living organisms protect the genome against external influences by recognizing and repairing damaged DNA. A common source of gene mutation is the oxidized guanine, which undergoes base excision repair through cleavage of the glycosidic bond between the ribose and the nucleobase of the lesion. We unravel the repair mechanism utilized by bacterial glycosylase, MutM, using quantum-chemical calculations involving more than 1000 atoms of the catalytic site. In contrast to the base-protonated pathway currently favored in the literature, we show that the initial protonation of the lesion's ribose paves the way for an almost barrier-free glycosidic cleavage. The combination of theoretical and experimental data provides further insight into the selectivity and discrimination of MutM's binding site toward various substrates.

Discovery and mutagenicity of a guanidinoformimine lesion as a new intermediate of the oxidative deoxyguanosine degradation pathway.

Oxidative degradation of DNA is a major mutagenic process. Reactive oxygen species (ROS) produced in the course of oxidative phosphorylation or by exogenous factors are known to attack preferentially deoxyguanosine. The latter decomposes to give mutagenic lesions, which under physiological conditions are efficiently repaired by specialized maintenance systems in the cell. Although many intermediates of the degradation pathway are today well-known, we report in this study the discovery of a new intermediate with an interesting guanidinoformimine structure. The structure elucidation of the new lesion was possible by using HPLC-MS techniques and organic synthesis. Finally we report the mutagenic potential of the new lesion in comparison to the known lesions imidazolone and oxazolone using primer extension and pyrosequencing experiments.

Improved synthesis and mutagenicity of oligonucleotides containing 5-hydroxymethylcytosine, 5-formylcytosine and 5-carboxylcytosine.

5-Formylcytosine (fC or (5-CHO)dC) and 5-carboxylcytosine (caC or (5-COOH)dC) have recently been identified as constituents of mammalian DNA. The nucleosides are formed from 5-methylcytosine (mC or (5-Me)dC) via 5-hydroxymethylcytosine (hmC or (5-HOMe)dC) and are possible intermediates of an active DNA demethylation process. Here we show efficient syntheses of phosphoramidites which enable the synthesis of DNA strands containing these cytosine modifications based on Pd(0)-catalyzed functionalization of 5-iododeoxycytidine. The first crystal structure of fC reveals the existence of an intramolecular H-bond between the exocyclic amine and the formyl group, which controls the conformation of the formyl substituent. Using a newly designed in vitro mutagenicity assay we show that fC and caC are only marginally mutagenic, which is a prerequisite for the bases to function as epigenetic control units.

8-Oxo-7,8-dihydroguanine in DNA does not constitute a barrier to transcription, but is converted into transcription-blocking damage by OGG1.

The common DNA base modification 8-oxo-7,8-dihydroguanine (8-oxo-G) affects the efficiency and fidelity of transcription. We constructed plasmid substrates carrying single 8-oxo-G residues, specifically positioned in the transcribed or the non-transcribed DNA strands, to investigate their effects on the expression of an EGFP reporter gene and to explore the role of base excision repair in the mechanism of transcription inhibition. We report that 8-oxo-G does not directly block transcription in cells, since a single 8-oxo-G in the transcribed DNA strand did not reduce the EGFP expression levels in repair-deficient (OGG1-null) mouse embryonic fibroblast cell lines. Rather, inhibition of transcription by 8-oxo-G fully depends on 8-oxoguanine DNA glycosylase (OGG1) and, at the same time, does not require the localization of the lesion in the transcribed DNA strand. We propose that the interruption of transcription is induced by base excision repair intermediates and, therefore, could be a common consequence of various DNA base modifications. Concordantly, the non-blocking DNA modification uracil was also found to inhibit transcription, but in an OGG1-independent manner.

Hysteroscopy in women with implantation failures after in vitro fertilization: findings and effect on subsequent pregnancy rates.

OBJECTIVE: To evaluate hysteroscopic findings and estimate the effect of hysteroscopy on achieving a pregnancy in women with a history of 2 implantation failures after in vitro fertilization (IVF). DESIGN: Prospective observational and matched case control study. DESIGN CLASSIFICATION: II-2. SETTING: Private assisted reproduction units. PATIENTS: A total of 1475 patients with a history of 2 consecutive implantation failures after IVF who had a hysteroscopy were studied; there were 414 matched pairs of women with a similar history who either had or did not have a hysteroscopy. INTERVENTIONS: Hysteroscopy (diagnostic or operative), IVF/intracytoplasmic sperm injection cycle. MEASUREMENTS AND MAIN RESULTS: Hysteroscopic findings, clinical pregnancy rate (CPR), and ongoing pregnancy rate (OPR) were measured. In all, 36.6% of the study population had abnormal hysteroscopic findings and 22.2% had unsuspected findings; women with abnormal hysteroscopic findings showed significantly increased CPR and increased OPR in a new IVF cycle compared with those with a normal hysteroscopy result. Women who had a hysteroscopy showed significantly increased CPR and OPR compared with matched control subjects who did not have the procedure. Hysteroscopy and appropriate therapy significantly increased the chances of achieving a subsequent clinical and ongoing pregnancy. CONCLUSION: Women with 2 implantation failures after IVF had a remarkably high possibility for unsuspected abnormalities seen at hysteroscopy. Hysteroscopy could serve as a positive prognostic factor for achieving a subsequent pregnancy.