Compared to conventional artificial neurons that produce dense and real-valued responses, biologically-inspired spiking neurons transmit sparse and binary information, which can also lead to energy-efficient implementations. Recent research has shown that spiking neural networks can be trained like standard recurrent neural networks using the surrogate gradient method. They have shown promising results on speech command recognition tasks. Using the same technique, we show that they are scalable to large vocabulary continuous speech recognition, where they are capable of replacing LSTMs in the encoder with only minor loss of performance. This suggests that they may be applicable to more involved sequence-to-sequence tasks. Moreover, in contrast to their recurrent non-spiking counterparts, they show robustness to exploding gradient problems without the need to use gates.

从机器学习的角度来看，当前的语音识别体系结构的表现非常出色，因此用户互动。这表明他们很好地模拟了人类生物系统。我们调查是否可以颠倒推论以提供对该生物系统的见解。特别是听力机制。使用SINCNET，我们确认端到端系统确实学习了众所周知的滤纸结构。但是，我们还表明，在学习结构中，更宽的带宽过滤器很重要。虽然可以通过初始化狭窄和宽带过滤器来获得一些好处，但生理上的限制表明，这种过滤器是在中脑而不是耳蜗中出现的。我们表明，必须修改标准的机器学习体系结构，以允许神经模拟此过程。

使用贝叶斯定理，我们得出了单位复发以及类似于前回去算法的后退递归。由此产生的贝叶斯复发单元可以在深度学习框架内集成为经常性神经网络，同时保留与隐藏的马尔可夫模型的直接对应关系的概率解释。尽管贡献主要是理论上的，但语音识别的实验表明，在最新的经常性架构结束时添加派生单元可以在可训练的参数方面以非常低的成本来提高性能。

Proteins play a central role in biology from immune recognition to brain activity. While major advances in machine learning have improved our ability to predict protein structure from sequence, determining protein function from structure remains a major challenge. Here, we introduce Holographic Convolutional Neural Network (H-CNN) for proteins, which is a physically motivated machine learning approach to model amino acid preferences in protein structures. H-CNN reflects physical interactions in a protein structure and recapitulates the functional information stored in evolutionary data. H-CNN accurately predicts the impact of mutations on protein function, including stability and binding of protein complexes. Our interpretable computational model for protein structure-function maps could guide design of novel proteins with desired function.

部署到现实世界的自主智能代理必须与对感官输入的对抗性攻击保持强大的态度。在加强学习中的现有工作集中于最小值扰动攻击，这些攻击最初是为了模仿计算机视觉中感知不变性的概念。在本文中，我们注意到，这种最小值扰动攻击可以由受害者琐碎地检测到，因为这些导致观察序列与受害者的行为不符。此外，许多现实世界中的代理商（例如物理机器人）通常在人类主管下运行，这些代理商不容易受到这种扰动攻击的影响。结果，我们建议专注于幻觉攻击，这是一种与受害者的世界模式一致的新型攻击形式。我们为这个新颖的攻击框架提供了正式的定义，在各种条件下探索了其特征，并得出结论，代理必须寻求现实主义反馈以对幻觉攻击具有强大的态度。

我们提出了Adios，这是一个用于自我监督学习的遮罩图像模型（MIM）框架，同时使用对抗性目标学习掩盖功能和图像编码器。对图像编码器进行了训练，以最大程度地减少原始图像的表示形式与蒙版图像的表示之间的距离。相反，掩蔽函数旨在最大化此距离。阿迪奥斯（Adios）始终改进有关各种任务和数据集的最先进的自我监督学习（SSL）方法 - 包括Imagenet100和STL10上的分类，CIFAR10/100上的转移学习，Flowers102和Inaturalist，以及鲁棒性在背景挑战中进行了评估（Xiao等，2021） - 同时产生语义意义的面具。与MAE，BEIT和IBOT等现代MIM模型不同，Adios不依赖视觉变压器的图像斑点令牌构造，并且可以用卷积的骨架来实现。我们进一步证明，与对流行MIM模型中使用的掩盖方案相比，阿迪奥斯学到的面具在改善SSL方法的表示方面更有效。

Multimodal VAEs seek to model the joint distribution over heterogeneous data (e.g.\ vision, language), whilst also capturing a shared representation across such modalities. Prior work has typically combined information from the modalities by reconciling idiosyncratic representations directly in the recognition model through explicit products, mixtures, or other such factorisations. Here we introduce a novel alternative, the MEME, that avoids such explicit combinations by repurposing semi-supervised VAEs to combine information between modalities implicitly through mutual supervision. This formulation naturally allows learning from partially-observed data where some modalities can be entirely missing -- something that most existing approaches either cannot handle, or do so to a limited extent. We demonstrate that MEME outperforms baselines on standard metrics across both partial and complete observation schemes on the MNIST-SVHN (image-image) and CUB (image-text) datasets. We also contrast the quality of the representations learnt by mutual supervision against standard approaches and observe interesting trends in its ability to capture relatedness between data.

We present a principled approach to incorporating labels in VAEs that captures the rich characteristic information associated with those labels. While prior work has typically conflated these by learning latent variables that directly correspond to label values, we argue this is contrary to the intended effect of supervision in VAEs-capturing rich label characteristics with the latents. For example, we may want to capture the characteristics of a face that make it look young, rather than just the age of the person. To this end, we develop the CCVAE, a novel VAE model and concomitant variational objective which captures label characteristics explicitly in the latent space, eschewing direct correspondences between label values and latents. Through judicious structuring of mappings between such characteristic latents and labels, we show that the CCVAE can effectively learn meaningful representations of the characteristics of interest across a variety of supervision schemes. In particular, we show that the CCVAE allows for more effective and more general interventions to be performed, such as smooth traversals within the characteristics for a given label, diverse conditional generation, and transferring characteristics across datapoints.

While the brain connectivity network can inform the understanding and diagnosis of developmental dyslexia, its cause-effect relationships have not yet enough been examined. Employing electroencephalography signals and band-limited white noise stimulus at 4.8 Hz (prosodic-syllabic frequency), we measure the phase Granger causalities among channels to identify differences between dyslexic learners and controls, thereby proposing a method to calculate directional connectivity. As causal relationships run in both directions, we explore three scenarios, namely channels' activity as sources, as sinks, and in total. Our proposed method can be used for both classification and exploratory analysis. In all scenarios, we find confirmation of the established right-lateralized Theta sampling network anomaly, in line with the temporal sampling framework's assumption of oscillatory differences in the Theta and Gamma bands. Further, we show that this anomaly primarily occurs in the causal relationships of channels acting as sinks, where it is significantly more pronounced than when only total activity is observed. In the sink scenario, our classifier obtains 0.84 and 0.88 accuracy and 0.87 and 0.93 AUC for the Theta and Gamma bands, respectively.

Differentiable Architecture Search (DARTS) has attracted considerable attention as a gradient-based Neural Architecture Search (NAS) method. Since the introduction of DARTS, there has been little work done on adapting the action space based on state-of-art architecture design principles for CNNs. In this work, we aim to address this gap by incrementally augmenting the DARTS search space with micro-design changes inspired by ConvNeXt and studying the trade-off between accuracy, evaluation layer count, and computational cost. To this end, we introduce the Pseudo-Inverted Bottleneck conv block intending to reduce the computational footprint of the inverted bottleneck block proposed in ConvNeXt. Our proposed architecture is much less sensitive to evaluation layer count and outperforms a DARTS network with similar size significantly, at layer counts as small as 2. Furthermore, with less layers, not only does it achieve higher accuracy with lower GMACs and parameter count, GradCAM comparisons show that our network is able to better detect distinctive features of target objects compared to DARTS.