Visual explanations from spiking neural networks using inter-spike intervals

By emulating biological features in brain, Spiking Neural Networks (SNNs) offer an energy-efficient alternative to conventional deep learning. To make SNNs ubiquitous, a ‘visual explanation’ technique for analysing and explaining the internal spike behavior of such temporal deep SNNs is crucial. Explaining SNNs visually will make the network more transparent giving the end-user a tool to understand how SNNs make temporal predictions and why they make a certain decision. In this paper, we propose a bio-plausible visual explanation tool for SNNs, called Spike Activation Map (SAM). SAM yields a heatmap (i.e., localization map) corresponding to each time-step of input data by highlighting neurons with short inter-spike interval activity. Interestingly, without the use of gradients and ground truth, SAM produces a temporal localization map highlighting the region of interest in an image attributed to an SNN’s prediction at each time-step. Overall, SAM outsets the beginning of a new research area ‘explainable neuromorphic computing’ that will ultimately allow end-users to establish appropriate trust in predictions from SNNs.

Consensus Patterns of a Set of Time Series via a Wavelet-Based Temporal Localization: Emphasizing the Utility over Point-Wise Averaging and Averaging under Dynamic Time Warping

Summarizing or averaging a sequential data set (i.e., a set of time series) can be comprehensively approached as a result of sophisticated computational tools. Averaging under Dynamic Time Warping (DTW) is one such tool that captures consensus patterns. DTW acts as a similarity measure between time series, and subsequently, an averaging method must be executed upon the behaviour of DTW. However, averaging under DTW somewhat neglects temporal aspect since it is on the search of similar appearances rather than stagnating on corresponding time-points. On the contrary, the mean series carrying point-wise averages provides only a weak consensus pattern as it may over-smooth important temporal variations. As a compromise, a pool of consensus series termed Ultimate Tamed Series (UTS) is studied here that adheres to temporal decomposition supported by the discrete Haar wavelet. We claim that UTS summarizes localized patterns, which would not be reachable via the series under DTW or the mean series. Neighbourhood of localization can be altered as a user can customize different levels of decomposition. In validation, comparisons are carried out with the series under DTW and the mean series via Euclidean distance and the distance resulted by DTW itself. Two sequential data sets are selected for this purpose from a standard repository.

Feeling the past: beyond causal content

Memories often come with a feeling of pastness. The events we remember strike us as having occurred in our past. What accounts for this feeling of pastness? In his recent book, Memory: A self-referential account, Jordi Fernández argues that the feeling of pastness cannot be grounded in an explicit representation of the pastness of the remembered event. Instead, he argues that the feeling of pastness is grounded in the self-referential causal content of memory. In this paper, I argue that this account falls short. The representation of causal origin does not by itself ground a feeling of pastness. Instead, I argue that we can salvage the temporal localization account of the feeling of pastness by describing a form of egocentric temporal representation that avoids Fernández’s criticisms.

Spatial and temporal localization of cell wall associated pili in Enterococcus faecalis

Enterococcus faecalis relies upon a number of cell wall-associated proteins for virulence. One virulence factor is the sortase-assembled endocarditis and biofilm associated pilus (Ebp), an important factor for biofilm formation in vitro and in vivo. The current paradigm for sortase-assembled pilus biogenesis in Gram-positive bacteria is that the pilus sortase covalently links pilus monomers prior to recognition, while the housekeeping sortase cleaves at the LPXTG motif within the terminal pilin subunit, and subsequently attaches assembled pilus fiber to the growing cell wall at sites of new cell wall synthesis. While the cell wall anchoring mechanism and polymerization of Ebp is well characterized, less is known about the spatial and temporal deposition of this protein on the cell surface. We followed the distribution of Ebp and peptidoglycan (PG) throughout the E. faecalis cell cycle via immunofluorescence microscopy and fluorescent D-amino acids (FDAA) staining. Surprisingly, cell surface Ebp did not co-localize with newly synthesized PG. Instead, surface-anchored Ebp was localized to the cell hemisphere but never at the septum where new cell wall is deposited. In addition, the older hemisphere of the E. faecalis diplococcus were completely saturated with Ebp, while Ebp appeared as two foci directly adjacent to the nascent septum in the newer hemisphere. A similar localization pattern was observed for another cell wall anchored substrate by sortase A, aggregation substance (AS), suggesting that this may be a general rule for all SrtA substrates in E. faecalis. When cell wall synthesis was inhibited by ramoplanin, an antibiotic that binds and sequesters lipid II cell wall precursors, new Ebp deposition at the cell surface was not disrupted. These data suggest an alternative paradigm for sortase substrate deposition in E. faecalis, in which Ebp are anchored directly onto un-crosslinked cell wall, independent of new PG synthesis.

Classification of Motor Imagery Electroencephalography Signals Based on Image Processing Method

In recent years, more and more frameworks have been applied to brain-computer interface technology, and electroencephalogram-based motor imagery (MI-EEG) is developing rapidly. However, it is still a challenge to improve the accuracy of MI-EEG classification. A deep learning framework termed IS-CBAM-convolutional neural network (CNN) is proposed to address the non-stationary nature, the temporal localization of excitation occurrence, and the frequency band distribution characteristics of the MI-EEG signal in this paper. First, according to the logically symmetrical relationship between the C3 and C4 channels, the result of the time-frequency image subtraction (IS) for the MI-EEG signal is used as the input of the classifier. It both reduces the redundancy and increases the feature differences of the input data. Second, the attention module is added to the classifier. A convolutional neural network is built as the base classifier, and information on the temporal location and frequency distribution of MI-EEG signal occurrences are adaptively extracted by introducing the Convolutional Block Attention Module (CBAM). This approach reduces irrelevant noise interference while increasing the robustness of the pattern. The performance of the framework was evaluated on BCI competition IV dataset 2b, where the mean accuracy reached 79.6%, and the average kappa value reached 0.592. The experimental results validate the feasibility of the framework and show the performance improvement of MI-EEG signal classification.

The Cartesian Folk Theater: People conceptualize consciousness as a spatio-temporally localized process in the human brain

The present research (total N = 2,057) tested whether people’s folk conception of consciousness aligns with the notion of a “Cartesian Theater” (Dennett, 1991). More precisely, we tested the hypotheses that people believe that consciousness happens in a single, confined area (vs. multiple dispersed areas) in the human brain, and that it (partly) happens after the brain finished analyzing all available information. Further, we investigated how these beliefs are related to participants’ neuroscientific knowledge as well as their reliance on intuition, and which rationale they use to explain their responses. Using a computer-administered drawing task, we found that participants located consciousness, but not unrelated neurological processes (Studies 1a & 1b) or unconscious thinking (Study 2) in a single, confined area in the prefrontal cortex, and that they considered most of the brain not involved in consciousness. Participants mostly relied on their intuitions when responding, and they were not affected by prior knowledge about the brain. Additionally, they considered the conscious experience of sensory stimuli to happen in a spatially more confined area than the corresponding computational analysis of these stimuli (Study 3). Furthermore, participants’ explicit beliefs about spatial and temporal localization of consciousness (i.e., consciousness happening after the computational analysis of sensory information is completed) are independent, yet positively correlated beliefs (Study 4). Using a more elaborate measure for temporal localization of conscious experience, our final study confirmed that people believe consciousness to partly happen even after information processing is done (Study 5).

Chimeric RNA:DNA Donorguide Improves HDR in vitro and in vivo

Introducing small genetic changes to study specific mutations or reverting clinical mutations to wild type has been an area of interest in precision genomics for several years. In fact, it has been found that nearly 90% of all human pathogenic mutations are caused by small genetic variations, and the methods to efficiently and precisely correct these errors are critically important. One common way to make these small DNA changes is to provide a single stranded DNA (ssDNA) donor containing the desired alteration together with a targeted double-strand break (DSB) at the genomic target. The donor is typically flanked by regions of homology that are often ~30-100bp in length to leverage the homology directed repair (HDR) pathway. Coupling a ssDNA donor with a CRISPR-Cas9 to produce a targeted DSB is one of the most streamlined approaches to introduce small changes. However, in many cell types this approach results in a low rate of incorporation of the desired alteration and has undesired imprecise repair at the 5' or 3' junction due to artifacts of the DNA repair process. We herein report a technology that couples the spatial temporal localization of an ssDNA repair template and leverages the nucleic acid components of the CRISPR-Cas9 system. We show that by direct fusion of an ssDNA template to the trans activating RNA (tracrRNA) to generate an RNA-DNA chimera, termed Donorguide, we recover precise integration of genetic alterations, with both increased integration rates and decreased imprecision at the 5' or 3' junctions relative to an ssODN donor in vitro in HEK293T cells as well as in vivo in zebrafish. Further, we show that this technology can be used to enhance gene conversion with other gene editing tools such as TALENs.