Exploiting nonlinear dendritic adaptive computation in training deep Spiking Neural Networks.

Inspired by the information transmission process in the brain, Spiking Neural Networks (SNNs) have gained considerable attention due to their event-driven nature. However, as the network structure grows complex, managing the spiking behavior within the network becomes challenging. Networks with excessively dense or sparse spikes fail to transmit sufficient information, inhibiting SNNs from exhibiting superior performance. Current SNNs linearly sum presynaptic information in postsynaptic neurons, overlooking the adaptive adjustment effect of dendrites on information processing. In this study, we introduce the Dendritic Spatial Gating Module (DSGM), which scales and translates the input, reducing the loss incurred when transforming the continuous membrane potential into discrete spikes. Simultaneously, by implementing the Dendritic Temporal Adjust Module (DTAM), dendrites assign different importance to inputs of different time steps, facilitating the establishment of the temporal dependency of spiking neurons and effectively integrating multi-step time information. The fusion of these two modules results in a more balanced spike representation within the network, significantly enhancing the neural network's performance. This approach has achieved state-of-the-art performance on static image datasets, including CIFAR10 and CIFAR100, as well as event datasets like DVS-CIFAR10, DVS-Gesture, and N-Caltech101. It also demonstrates competitive performance compared to the current state-of-the-art on the ImageNet dataset.

Brain-inspired neural circuit evolution for spiking neural networks.

In biological neural systems, different neurons are capable of self-organizing to form different neural circuits for achieving a variety of cognitive functions. However, the current design paradigm of spiking neural networks is based on structures derived from deep learning. Such structures are dominated by feedforward connections without taking into account different types of neurons, which significantly prevent spiking neural networks from realizing their potential on complex tasks. It remains an open challenge to apply the rich dynamical properties of biological neural circuits to model the structure of current spiking neural networks. This paper provides a more biologically plausible evolutionary space by combining feedforward and feedback connections with excitatory and inhibitory neurons. We exploit the local spiking behavior of neurons to adaptively evolve neural circuits such as forward excitation, forward inhibition, feedback inhibition, and lateral inhibition by the local law of spike-timing-dependent plasticity and update the synaptic weights in combination with the global error signals. By using the evolved neural circuits, we construct spiking neural networks for image classification and reinforcement learning tasks. Using the brain-inspired Neural circuit Evolution strategy (NeuEvo) with rich neural circuit types, the evolved spiking neural network greatly enhances capability on perception and reinforcement learning tasks. NeuEvo achieves state-of-the-art performance on CIFAR10, DVS-CIFAR10, DVS-Gesture, and N-Caltech101 datasets and achieves advanced performance on ImageNet. Combined with on-policy and off-policy deep reinforcement learning algorithms, it achieves comparable performance with artificial neural networks. The evolved spiking neural circuits lay the foundation for the evolution of complex networks with functions.

BrainCog: A spiking neural network based, brain-inspired cognitive intelligence engine for brain-inspired AI and brain simulation.

Spiking neural networks (SNNs) serve as a promising computational framework for integrating insights from the brain into artificial intelligence (AI). Existing software infrastructures based on SNNs exclusively support brain simulation or brain-inspired AI, but not both simultaneously. To decode the nature of biological intelligence and create AI, we present the brain-inspired cognitive intelligence engine (BrainCog). This SNN-based platform provides essential infrastructure support for developing brain-inspired AI and brain simulation. BrainCog integrates different biological neurons, encoding strategies, learning rules, brain areas, and hardware-software co-design as essential components. Leveraging these user-friendly components, BrainCog incorporates various cognitive functions, including perception and learning, decision-making, knowledge representation and reasoning, motor control, social cognition, and brain structure and function simulations across multiple scales. BORN is an AI engine developed by BrainCog, showcasing seamless integration of BrainCog's components and cognitive functions to build advanced AI models and applications.

[Fingerprint of sophorolipids based on ultra-high performance liquid chromatography-charged aerosol detection].

Sophorolipids are secondary metabolites produced during fermentation by nonpathogenic yeasts. These molecules are amphiphilic and consist of a hydrophilic sophora sugar moiety and a hydrophobic hydroxylated fatty acid. Based on their degree of esterification, sophorolipids can be divided into the acid and lactone types. Sophorolipids are highly promising biosurfactants with good antibacterial, antiviral, and other biological activities. Moreover, they are characterized by mildness, low toxicity, and environmental friendliness. However, their composition is quite complex, and effective methods for their quality evaluation are lacking. Since sophorolipids do not absorb ultraviolet (UV) light, common UV detectors are unsuitable for fingerprint establishment. In this study, we first selected a charged aerosol detector (CAD) to establish the ultra-high performance liquid chromatography (UHPLC) fingerprint of sophorolipids. The detector had high sensitivity, good reproducibility, and excellent suitability for the detection of substances with no or weak ultraviolet absorption. We then evaluated the similarities between 17 batches of sophorolipid samples. The samples were extracted by ultrasound for 10 min in 80% ethanol aqueous solution at a liquid-solid ratio of 10∶1 (mL/g) and then separated on a Thermo Fisher Scientific Hypersil Gold chromatographic column (150 mm×2.1 mm, 1.9 µm). Separation was performed using acetonitrile-0.01% (v/v) formic acid aqueous solution as the mobile phase via gradient elution. The flow rate was 0.2 mL/min, and the column temperature was 40 ℃. The CAD was used under the following conditions: power function of 1.0, data rate of 5 Hz, filter constant of 3.6, and evaporation temperature of 45 ℃. The chromatograms and retention times of the sophorolipids were compared, and 16 common peaks with strong responses, good resolutions, and stable retention times were selected as characteristic peaks. Oleic acid was chosen as the reference peak because it achieved good separation and a strong chromatographic response in all batches of samples. UHPLC-quadrupole time-of-flight mass spectrometry (UHPLC-QTOF-MS) was used to identify chromatographic peaks in the sophorolipid fingerprints. The results were combined with the retention time rule of the sophorolipids, leading to their identification based on matching with the results of the primary database, the precise relative molecular mass and fragmentation rule of secondary fragments, a self-built database, and the PubChem database. Sixteen compounds were identified, including eight acid sophorolipids, six lactone sophorolipids, and two aliphatic acids. The results of precision, repeatability, and 24 h stability tests indicated that the relative standard deviations (RSDs) of the retention times and peak areas of the 15 characteristic peaks relative to the control peak (oleic acid) were less than 3.0% (n=6). Seventeen batches of sophorolipid samples were analyzed, and the similarity values of all fingerprints were found to be 0.965 or higher. Little differences in chemical composition were observed among the different batches of sophorolipid samples, and the quality of the sophorolipids was relatively consistent. The fingerprint established in this study is stable and reliable; it can be used for the quality evaluation of sophorolipids and lays a solid foundation for future research on production technology and the development and utilization of sophorolipids. The successful application of a universal CAD to the fingerprint establishment of sophorolipids also provides a reliable solution for the fingerprint establishment of substances with no or weak ultraviolet absorption.

Backpropagation with biologically plausible spatiotemporal adjustment for training deep spiking neural networks.

The spiking neural network (SNN) mimics the information-processing operation in the human brain. Directly applying backpropagation to the training of the SNN still has a performance gap compared with traditional deep neural networks. To address the problem, we propose a biologically plausible spatial adjustment that rethinks the relationship between membrane potential and spikes and realizes a reasonable adjustment of gradients to different time steps. It precisely controls the backpropagation of the error along the spatial dimension. Secondly, we propose a biologically plausible temporal adjustment to make the error propagate across the spikes in the temporal dimension, which overcomes the problem of the temporal dependency within a single spike period of traditional spiking neurons. We have verified our algorithm on several datasets, and the experimental results have shown that our algorithm greatly reduces network latency and energy consumption while also improving network performance.

An integrated device of electrodialytic membrane suppressor and charge detector for ion chromatography.

An integrated device of electrodialytic membrane suppressor and charge detector (CHD) for ion chromatography (IC) is described, aiming to simplify system complexity and reduce possible extra-column dispersion as well. The device is a flow-through design consisting of five chambers isolated by ion exchange membranes and ionic screens. Two independent electric sources are used to respectively supply CHD and suppressor sections of the device, and a common electrode serves both the cathode of CHD and the anode of suppressor simultaneously. The integrated device has similar performance as a separate suppressor or a CHD while its dead volume and dispersion are reduced ∼18% and ∼37% compared with the combination of a CHD and a suppressor. To our knowledge, this is the first description of such an integrated device with dual functionalities of suppression and CHD.

Fabrication and evaluation of an electrodialytic carbonate eluent generator for ion chromatography.

An electrodialytic potassium carbonate eluent generator and its associated potassium bicarbonate eluent generator have been fabricated for ion chromatography (IC). The device can withstand high backpressure up to ∼32MPa and no observable leakage under such pressure is found during 2h. In the range of 0-13.7mM, potassium carbonate concentration can be generated linearly with the applied current with a slope that is essentially Faradaic. At least 10mM potassium carbonate can be online changed into 10mM potassium bicarbonate via a potassium bicarbonate eluent generator, which offers an easy way to manipulate the separation selectivity. When coupled with IC system, the device demonstrated good reproducibility indicated by less than 0.52% of the relative standard deviation of the retention times.

A hyperbranched polyethylenimine functionalized stationary phase for hydrophilic interaction liquid chromatography.

A hyperbranched stationary phase for hydrophilic interaction liquid chromatography (HILIC) has been prepared by grafting polyethylenimine (PEI) onto silica gel (termed as PEI-Sil). Rich primary, secondary, and tertiary amino groups associated with PEI render its good hydrophility. More importantly, the hyperbranched structure of PEI molecule is greatly helpful in improving interaction with polar analytes. For several kinds of model polar compounds, including organic acids, nucleosides, nucleic acid bases, amino acids, cephalosporins, and non-reducing sugars, PEI-Sil demonstrated excellent separation performance in terms of running stability, reproducibility, and separation efficiency (e.g., plate count ~74,000/m). In addition, PEI-Sil also exhibited much better separation selectivity toward inorganic anions when operated in the mode of ion chromatography relative to a commercial amino propyl-bonded column.

Click aspartic acid as H ILIC SPE material for selective enrichment of N-linked glycopeptides.

A novel hydrophilic interaction chromatography (HILIC) material was developed via clicking aspartic acid onto silica gel (termed as Click AA). The material demonstrated highly hydrophilic property and was used as solid phase extraction sorbent for selective enrichment of glycopeptides. By optimization of extraction conditions, Click AA exhibited high selectivity to glycopeptides with various peptide length and different types of glycans, which was much superior to similar commercial products. The application of Click AA to simulated proteomic samples further proved its specialty toward glycopeptides.

Simple preparation of solid-phase microextraction fiber with cation exchange capacities using poly(butadiene-maleic acid).

A simply way was proposed to prepare solid-phase microextraction (SPME) fiber with cation-exchange functional groups by the thermally initiated radical polymerization of poly(butadiene-maleic acid) (PBMA) copolymer onto a silica capillary. The capacity of the fiber coating could be easily controlled by fabricating successive layers of PBMA. The performance of the fiber combined with ion chromatography was evaluated by choosing Mg(2+) and Ca(2+) as model analytes; ∼13 and ∼51-fold enrichment factors for Mg(2+) and Ca(2+) were obtained, respectively.

A novel amide stationary phase for hydrophilic interaction liquid chromatography and ion chromatography.

A novel amide stationary phase (ASP) for hydrophilic interaction liquid chromatography (HILIC) has been prepared via the Click chemistry method. It was based on the strategy that the amino group of Asparagine was easily transferred to the corresponding azido group and then clicked onto terminal alkyne-silica gel in the presence of Cu(I)-based catalyst. For the tested polar compounds including nucleosides and nucleic acid bases, ASP-based column has demonstrated good performance in terms of separation efficiency and column stability, and the retention mechanism was found to match well the typical HILIC retention. In addition, the ASP described here showed much better selectivity in separation of inorganic anions under ion chromatography mode relative to other kinds of commercial ASP.

Comprehensive HILIC × RPLC with mass spectrometry detection for the analysis of saponins in Panax notoginseng.

A comprehensive off-line two-dimensional liquid chromatography (2D-LC) method coupling hydrophilic interaction liquid chromatography (HILIC) and reversed phase liquid chromatography (RPLC) was developed in this study to detect as many saponins as possible in extracts of Panax notoginseng. The orthogonality of the 2D HILIC × RPLC was up to 81%, and the peak capacity was 10200. In total, 224 saponins were found, and some of them were trace amounts. Besides, a screening table designed by adding molecular weights of possible aglycones and sugars was constructed to help rapidly characterize the saponins using MS information. Unfortunately, the structure of saponins could not be identified by using only MS information.

Adaptive vector quantization with codebook updating based on locality and history.

In this paper, we propose two techniques that are applicable to any adaptive vector quantization (AVQ) systems. The first one is called the locality-based codebook updating: when performing a codebook updating, we update the operational codebook using not only the current input vector but also the codewords at all positions within a selected neighboring area (called the locality), while the operational codebook is organized in a "cache" manner. This technique is rationalized by the high correlation cross neighboring vectors that facilitates a more efficient coding of the indices of the codewords chosen from the codebook. The second technique is called the history aid, which makes use of the information of previously coded vectors to quantize the current input vector if it is used to update the operational codebook. A more effective AVQ system is obtained by combining together the history aid and the locality-based updating. Extensive simulations are carried out to demonstrate the improved results achieved by our AVQ systems. Particularly, when the operational codebook size is relatively small, the improvement over a benchmark AVQ system--the generalized threshold replenishment (GTR)--is drastic. For example, when the size is 32, testing on a nonstationary signal (containing frames from different video sequences, ordered in the concatenating or interleaving format) shows that the combination of history aid and locality-based updating offers more than 4 dB gain over GTR at 0.5 bpp.