Recent analyses of neural networks with shaped activations (i.e. the activation function is scaled as the network size grows) have led to scaling limits described by differential equations. However, these results do not a priori tell us anything about "ordinary" unshaped networks, where the activation is unchanged as the network size grows. In this article, we find similar differential equation based asymptotic characterization for two types of unshaped networks. Firstly, we show that the following two architectures converge to the same infinite-depth-and-width limit at initialization: (i) a fully connected ResNet with a $d^{-1/2}$ factor on the residual branch, where $d$ is the network depth. (ii) a multilayer perceptron (MLP) with depth $d \ll$ width $n$ and shaped ReLU activation at rate $d^{-1/2}$. Secondly, for an unshaped MLP at initialization, we derive the first order asymptotic correction to the layerwise correlation. In particular, if $\rho_\ell$ is the correlation at layer $\ell$, then $q_t = \ell^2 (1 - \rho_\ell)$ with $t = \frac{\ell}{n}$ converges to an SDE with a singularity at $t=0$. These results together provide a connection between shaped and unshaped network architectures, and opens up the possibility of studying the effect of normalization methods and how it connects with shaping activation functions.

Neural networks with wide layers have attracted significant attention due to their equivalence to Gaussian processes, enabling perfect fitting of training data while maintaining generalization performance, known as benign overfitting. However, existing results mainly focus on shallow or finite-depth networks, necessitating a comprehensive analysis of wide neural networks with infinite-depth layers, such as neural ordinary differential equations (ODEs) and deep equilibrium models (DEQs). In this paper, we specifically investigate the deep equilibrium model (DEQ), an infinite-depth neural network with shared weight matrices across layers. Our analysis reveals that as the width of DEQ layers approaches infinity, it converges to a Gaussian process, establishing what is known as the Neural Network and Gaussian Process (NNGP) correspondence. Remarkably, this convergence holds even when the limits of depth and width are interchanged, which is not observed in typical infinite-depth Multilayer Perceptron (MLP) networks. Furthermore, we demonstrate that the associated Gaussian vector remains non-degenerate for any pairwise distinct input data, ensuring a strictly positive smallest eigenvalue of the corresponding kernel matrix using the NNGP kernel. These findings serve as fundamental elements for studying the training and generalization of DEQs, laying the groundwork for future research in this area.

Track one of CTI competition is on click-through rate (CTR) prediction. The dataset contains millions of records and each field-wise feature in a record consists of hashed integers for privacy. For this task, the keys of network-based methods might be type-wise feature extraction and information fusion across different fields. Multi-layer perceptrons (MLPs) are able to extract field feature, but could not efficiently fuse features. Motivated by the natural fusion characteristic of cross attention and the efficiency of transformer-based structures, we propose simple plug-in mixers for field/type-wise feature fusion, and thus construct an field&type-wise ensemble model, namely EMOFM (Ensemble MLP mOdel with Feature-based Mixers). In the experiments, the proposed model is evaluated on the dataset, the optimization process is visualized and ablation studies are explored. It is shown that EMOFM outperforms compared baselines. In the end, we discuss on future work. WARNING: The comparison might not be fair enough since the proposed method is designed for this data in particular while compared methods are not. For example, EMOFM especially takes different types of interactions into consideration while others do not. Anyway, we do hope that the ideas inside our method could help other developers/learners/researchers/thinkers and so on.

Autoregressive~(AR) generation almost dominates sequence generation for its efficacy. Recently, non-autoregressive~(NAR) generation gains increasing popularity for its efficiency and growing efficacy. However, its efficiency is still bottlenecked by quadratic complexity in sequence lengths, which is prohibitive for scaling to long sequence generation and few works have been done to mitigate this problem. In this paper, we propose a novel MLP variant, \textbf{A}ttentive \textbf{M}ulti-\textbf{L}ayer \textbf{P}erceptron~(AMLP), to produce a generation model with linear time and space complexity. Different from classic MLP with static and learnable projection matrices, AMLP leverages adaptive projections computed from inputs in an attentive mode. The sample-aware adaptive projections enable communications among tokens in a sequence, and model the measurement between the query and key space. Furthermore, we marry AMLP with popular NAR models, deriving a highly efficient NAR-AMLP architecture with linear time and space complexity. Empirical results show that such marriage architecture surpasses competitive efficient NAR models, by a significant margin on text-to-speech synthesis and machine translation. We also test AMLP's self- and cross-attention ability separately with extensive ablation experiments, and find them comparable or even superior to the other efficient models. The efficiency analysis further shows that AMLP extremely reduces the memory cost against vanilla non-autoregressive models for long sequences.

This study presents a novel multimodal fusion model for three-dimensional mineral prospectivity mapping (3D MPM), effectively integrating structural and fluid information through a deep network architecture. Leveraging Convolutional Neural Networks (CNN) and Multilayer Perceptrons (MLP), the model employs canonical correlation analysis (CCA) to align and fuse multimodal features. Rigorous evaluation on the Jiaojia gold deposit dataset demonstrates the model's superior performance in distinguishing ore-bearing instances and predicting mineral prospectivity, outperforming other models in result analyses. Ablation studies further reveal the benefits of joint feature utilization and CCA incorporation. This research not only advances mineral prospectivity modeling but also highlights the pivotal role of data integration and feature alignment for enhanced exploration decision-making.

In recent years, legged robots based on deep reinforcement learning have made remarkable progress. Quadruped robots have demonstrated the ability to complete challenging tasks in complex environments and have been deployed in real-world scenarios to assist humans. Simultaneously, bipedal and humanoid robots have achieved breakthroughs in various demanding tasks. Current reinforcement learning methods can utilize diverse robot bodies and historical information to perform actions. However, prior research has not emphasized the speed and energy consumption of network inference, as well as the biological significance of the neural networks themselves. Most of the networks employed are traditional artificial neural networks that utilize multilayer perceptrons (MLP). In this paper, we successfully apply a novel Spiking Neural Network (SNN) to process legged robots, achieving outstanding results across a range of simulated terrains. SNN holds a natural advantage over traditional neural networks in terms of inference speed and energy consumption, and their pulse-form processing of body perception signals offers improved biological interpretability. To the best of our knowledge, this is the first work to implement SNN in legged robots.

This paper proposes an adjoint-equivariant neural network that takes Lie algebra data as input. Various types of equivariant neural networks have been proposed in the literature, which treat the input data as elements in a vector space carrying certain types of transformations. In comparison, we aim to process inputs that are transformations between vector spaces. The change of basis on transformation is described by conjugations, inducing the adjoint-equivariance relationship that our model is designed to capture. Leveraging the invariance property of the Killing form, the proposed network is a general framework that works for arbitrary semisimple Lie algebras. Our network possesses a simple structure that can be viewed as a Lie algebraic generalization of a multi-layer perceptron (MLP). This work extends the application of equivariant feature learning. Respecting the symmetry in data is essential for deep learning models to understand the underlying objects.

Recently, it has been shown that for offline deep reinforcement learning (DRL), pre-training Decision Transformer with a large language corpus can improve downstream performance (Reid et al., 2022). A natural question to ask is whether this performance gain can only be achieved with language pre-training, or can be achieved with simpler pre-training schemes which do not involve language. In this paper, we first show that language is not essential for improved performance, and indeed pre-training with synthetic IID data for a small number of updates can match the performance gains from pre-training with a large language corpus; moreover, pre-training with data generated by a one-step Markov chain can further improve the performance. Inspired by these experimental results, we then consider pre-training Conservative Q-Learning (CQL), a popular offline DRL algorithm, which is Q-learning-based and typically employs a Multi-Layer Perceptron (MLP) backbone. Surprisingly, pre-training with simple synthetic data for a small number of updates can also improve CQL, providing consistent performance improvement on D4RL Gym locomotion datasets. The results of this paper not only illustrate the importance of pre-training for offline DRL but also show that the pre-training data can be synthetic and generated with remarkably simple mechanisms. It is well-known that pre-training can provide significant boosts in performance and robustness for downstream tasks, both for Natural Language Processing (NLP) and Computer Vision (CV). Recently, in the field of Deep Reinforcement Learning (DRL), research on pre-training is also becoming increasingly popular. An important step in the direction of pre-training DRL models is the recent paper by Reid et al. (2022), which showed that for Decision Transformer (Chen et al., 2021), pretraining with the Wikipedia corpus can significantly improve the performance of the downstream offline RL task. Reid et al. (2022) further showed that pre-training on predicting pixel sequences can hurt performance. The authors state that their results indicate "a foreseeable future where everyone should use a pre-trained language model for offline RL".

Model binarization can significantly compress model size, reduce energy consumption, and accelerate inference through efficient bit-wise operations. Although binarizing convolutional neural networks have been extensively studied, there is little work on exploring binarization of vision Transformers which underpin most recent breakthroughs in visual recognition. To this end, we propose to solve two fundamental challenges to push the horizon of Binary Vision Transformers (BiViT). First, the traditional binary method does not take the long-tailed distribution of softmax attention into consideration, bringing large binarization errors in the attention module. To solve this, we propose Softmax-aware Binarization, which dynamically adapts to the data distribution and reduces the error caused by binarization. Second, to better preserve the information of the pretrained model and restore accuracy, we propose a Cross-layer Binarization scheme that decouples the binarization of self-attention and multi-layer perceptrons (MLPs), and Parameterized Weight Scales which introduce learnable scaling factors for weight binarization. Overall, our method performs favorably against state-of-the-arts by 19.8% on the TinyImageNet dataset. On ImageNet, our BiViT achieves a competitive 75.6% Top-1 accuracy over Swin-S model. Additionally, on COCO object detection, our method achieves an mAP of 40.8 with a Swin-T backbone over Cascade Mask R-CNN framework.

Point convolution operations rely on different embedding mechanisms to encode the neighborhood information of each point in order to detect patterns in 3D space. However, as convolutions are usually evaluated as a whole, not much work has been done to investigate which is the ideal mechanism to encode such neighborhood information. In this paper, we provide the first extensive study that analyzes such Point Neighborhood Embeddings (PNE) alone in a controlled experimental setup. From our experiments, we derive a set of recommendations for PNE that can help to improve future designs of neural network architectures for point clouds. Our most surprising finding shows that the most commonly used embedding based on a Multi-layer Perceptron (MLP) with ReLU activation functions provides the lowest performance among all embeddings, even being surpassed on some tasks by a simple linear combination of the point coordinates. Additionally, we show that a neural network architecture using simple convolutions based on such embeddings is able to achieve state-of-the-art results on several tasks, outperforming recent and more complex operations. Lastly, we show that these findings extrapolate to other more complex convolution operations, where we show how following our recommendations we are able to improve recent state-of-the-art architectures. In computer vision, point clouds are one of the most commonly used representations to process and store 3D data. This is because point clouds are a compact representation and most 3D acquisition hardware produces point clouds as their output. In the past years, the advances in 3D acquisition hardware, and therefore the number of available point clouds, allowed the development of data-driven methods to solve different 3D scene understanding problems. In particular, the pioneering work of Qi et al. (2017a) opened the door to new neural network architectures which were able to process point clouds directly.