Deep ReLU Networks Have Surprisingly Simple Polytopes

A ReLU network is a piecewise linear function over polytopes. Figuring out the properties of such polytopes is of fundamental importance for the research and development of neural networks. So far, either theoretical or empirical studies on polytopes only stay at the level of counting their number, which is far from a complete characterization of polytopes. To upgrade the characterization to a new level, here we propose to study the shapes of polytopes via the number of simplices obtained by triangulating the polytope. Then, by computing and analyzing the histogram of simplices across polytopes, we find that a ReLU network has relatively simple polytopes under both initialization and gradient descent, although these polytopes theoretically can be rather diverse and complicated. This finding can be appreciated as a novel implicit bias. Next, we use nontrivial combinatorial derivation to theoretically explain why adding depth does not create a more complicated polytope by bounding the average number of faces of polytopes with a function of the dimensionality. Our results concretely reveal what kind of simple functions a network learns and its space partition property. Also, by characterizing the shape of polytopes, the number of simplices be a leverage for other problems, \textit{e.g.}, serving as a generic functional complexity measure to explain the power of popular shortcut networks such as ResNet and analyzing the impact of different regularization strategies on a network's space partition.

A data-driven rutting depth short-time prediction model with metaheuristic optimization for asphalt pavements based on RIOHTrack

Rutting of asphalt pavements is a crucial design criterion in various pavement design guides. A good road transportation base can provide security for the transportation of oil and gas in road transportation. This study attempts to develop a robust artificial intelligence model to estimate different asphalt pavements' rutting depth clips, temperature, and load axes as primary characteristics. The experiment data were obtained from 19 asphalt pavements with different crude oil sources on a 2.038 km long full-scale field accelerated pavement test track (RIOHTrack, Road Track Institute) in Tongzhou, Beijing. In addition, this paper also proposes to build complex networks with different pavement rutting depths through complex network methods and the Louvain algorithm for community detection. The most critical structural elements can be selected from different asphalt pavement rutting data, and similar structural elements can be found. An extreme learning machine algorithm with residual correction (RELM) is designed and optimized using an independent adaptive particle swarm algorithm. The experimental results of the proposed method are compared with several classical machine learning algorithms, with predictions of Average Root Mean Squared Error, Average Mean Absolute Error, and Average Mean Absolute Percentage Error for 19 asphalt pavements reaching 1.742, 1.363, and 1.94\% respectively. The experiments demonstrate that the RELM algorithm has an advantage over classical machine learning methods in dealing with non-linear problems in road engineering. Notably, the method ensures the adaptation of the simulated environment to different levels of abstraction through the cognitive analysis of the production environment parameters.

Towards Understanding Feature Learning in Out-of-Distribution Generalization

A common explanation for the failure of out-of-distribution (OOD) generalization is that the model trained with empirical risk minimization (ERM) learns spurious features instead of the desired invariant features. However, several recent studies challenged this explanation and found that deep networks may have already learned sufficiently good features for OOD generalization. The debate extends to the in-distribution and OOD performance correlations along with training or fine-tuning neural nets across a variety of OOD generalization tasks. To understand these seemingly contradicting phenomena, we conduct a theoretical investigation and find that ERM essentially learns both spurious features and invariant features. On the other hand, the quality of learned features during ERM pre-training significantly affects the final OOD performance, as OOD objectives rarely learn new features. Failing to capture all the underlying useful features during pre-training will further limit the final OOD performance. To remedy the issue, we propose Feature Augmented Training (FAT ), to enforce the model to learn all useful features by retaining the already learned features and augmenting new ones by multiple rounds. In each round, the retention and augmentation operations are performed on different subsets of the training data that capture distinct features. Extensive experiments show that FAT effectively learns richer features and consistently improves the OOD performance when applied to various objectives.

Large Language Models as Master Key: Unlocking the Secrets of Materials Science with GPT

The amount of data has growing significance in exploring cutting-edge materials and a number of datasets have been generated either by hand or automated approaches. However, the materials science field struggles to effectively utilize the abundance of data, especially in applied disciplines where materials are evaluated based on device performance rather than their properties. This article presents a new natural language processing (NLP) task called structured information inference (SII) to address the complexities of information extraction at the device level in materials science. We accomplished this task by tuning GPT-3 on an existing perovskite solar cell FAIR (Findable, Accessible, Interoperable, Reusable) dataset with 91.8% F1-score and extended the dataset with data published since its release. The produced data is formatted and normalized, enabling its direct utilization as input in subsequent data analysis. This feature empowers materials scientists to develop models by selecting high-quality review articles within their domain. Additionally, we designed experiments to predict the electrical performance of solar cells and design materials or devices with targeted parameters using large language models (LLMs). Our results demonstrate comparable performance to traditional machine learning methods without feature selection, highlighting the potential of LLMs to acquire scientific knowledge and design new materials akin to materials scientists.

CASP-Net: Rethinking Video Saliency Prediction from an Audio-VisualConsistency Perceptual Perspective

Incorporating the audio stream enables Video Saliency Prediction (VSP) to imitate the selective attention mechanism of human brain. By focusing on the benefits of joint auditory and visual information, most VSP methods are capable of exploiting semantic correlation between vision and audio modalities but ignoring the negative effects due to the temporal inconsistency of audio-visual intrinsics. Inspired by the biological inconsistency-correction within multi-sensory information, in this study, a consistency-aware audio-visual saliency prediction network (CASP-Net) is proposed, which takes a comprehensive consideration of the audio-visual semantic interaction and consistent perception. In addition a two-stream encoder for elegant association between video frames and corresponding sound source, a novel consistency-aware predictive coding is also designed to improve the consistency within audio and visual representations iteratively. To further aggregate the multi-scale audio-visual information, a saliency decoder is introduced for the final saliency map generation. Substantial experiments demonstrate that the proposed CASP-Net outperforms the other state-of-the-art methods on six challenging audio-visual eye-tracking datasets. For a demo of our system please see our project webpage.

Generic-to-Specific Distillation of Masked Autoencoders

Large vision Transformers (ViTs) driven by self-supervised pre-training mechanisms achieved unprecedented progress. Lightweight ViT models limited by the model capacity, however, benefit little from those pre-training mechanisms. Knowledge distillation defines a paradigm to transfer representations from large (teacher) models to small (student) ones. However, the conventional single-stage distillation easily gets stuck on task-specific transfer, failing to retain the task-agnostic knowledge crucial for model generalization. In this study, we propose generic-to-specific distillation (G2SD), to tap the potential of small ViT models under the supervision of large models pre-trained by masked autoencoders. In generic distillation, decoder of the small model is encouraged to align feature predictions with hidden representations of the large model, so that task-agnostic knowledge can be transferred. In specific distillation, predictions of the small model are constrained to be consistent with those of the large model, to transfer task-specific features which guarantee task performance. With G2SD, the vanilla ViT-Small model respectively achieves 98.7%, 98.1% and 99.3% the performance of its teacher (ViT-Base) for image classification, object detection, and semantic segmentation, setting a solid baseline for two-stage vision distillation. Code will be available at https://github.com/pengzhiliang/G2SD.

DMMG: Dual Min-Max Games for Self-Supervised Skeleton-Based Action Recognition

In this work, we propose a new Dual Min-Max Games (DMMG) based self-supervised skeleton action recognition method by augmenting unlabeled data in a contrastive learning framework. Our DMMG consists of a viewpoint variation min-max game and an edge perturbation min-max game. These two min-max games adopt an adversarial paradigm to perform data augmentation on the skeleton sequences and graph-structured body joints, respectively. Our viewpoint variation min-max game focuses on constructing various hard contrastive pairs by generating skeleton sequences from various viewpoints. These hard contrastive pairs help our model learn representative action features, thus facilitating model transfer to downstream tasks. Moreover, our edge perturbation min-max game specializes in building diverse hard contrastive samples through perturbing connectivity strength among graph-based body joints. The connectivity-strength varying contrastive pairs enable the model to capture minimal sufficient information of different actions, such as representative gestures for an action while preventing the model from overfitting. By fully exploiting the proposed DMMG, we can generate sufficient challenging contrastive pairs and thus achieve discriminative action feature representations from unlabeled skeleton data in a self-supervised manner. Extensive experiments demonstrate that our method achieves superior results under various evaluation protocols on widely-used NTU-RGB+D and NTU120-RGB+D datasets.

Communication under Mixed Gaussian-Impulsive Channel: An End-to-End Framework

In many communication scenarios, the communication signals simultaneously suffer from white Gaussian noise (WGN) and non-Gaussian impulsive noise (IN), i.e., mixed Gaussian-impulsive noise (MGIN). Under MGIN channel, classical communication signal schemes and corresponding detection methods usually can not achieve desirable performance as they are optimized with respect to WGN. Moreover, as the widely adopted IN model has no analytical and general closed-form expression of probability density function (PDF), it is extremely hard to obtain optimal communication signal and corresponding detection schemes based on classical stochastic signal processing theory. To circumvent these difficulties, we propose a data-driven end-to-end framework to address the communication signal design and detection under MGIN channel in this paper. In this proposed framework, a channel noise simulator (CNS) is elaborately designed based on an improved generative adversarial net (GAN) to simulate the MGIN without requirement of any analytical PDF. Meanwhile, a multi-level wavelet convolutional neural network (MWCNN) based preprocessing network is used to mitigate the negative effect of outliers due to the IN. Compared with conventional approaches and existing end-to-end systems, extensive simulation results verify that our proposed novel end-to-end communication system can achieve better performance in terms of bit-error rate (BER) under MGIN environments.

TCFimt: Temporal Counterfactual Forecasting from Individual Multiple Treatment Perspective

Determining causal effects of temporal multi-intervention assists decision-making. Restricted by time-varying bias, selection bias, and interactions of multiple interventions, the disentanglement and estimation of multiple treatment effects from individual temporal data is still rare. To tackle these challenges, we propose a comprehensive framework of temporal counterfactual forecasting from an individual multiple treatment perspective (TCFimt). TCFimt constructs adversarial tasks in a seq2seq framework to alleviate selection and time-varying bias and designs a contrastive learning-based block to decouple a mixed treatment effect into separated main treatment effects and causal interactions which further improves estimation accuracy. Through implementing experiments on two real-world datasets from distinct fields, the proposed method shows satisfactory performance in predicting future outcomes with specific treatments and in choosing optimal treatment type and timing than state-of-the-art methods.

Can Single-Pass Contrastive Learning Work for Both Homophilic and Heterophilic Graph?

Existing graph contrastive learning (GCL) typically requires two forward pass for a single instance to construct the contrastive loss. Despite its remarkable success, it is unclear whether such a dual-pass design is (theoretically) necessary. Besides, the empirical results are hitherto limited to the homophilic graph benchmarks. Then a natural question arises: Can we design a method that works for both homophilic and heterophilic graphs with a performance guarantee? To answer this, we analyze the concentration property of features obtained by neighborhood aggregation on both homophilic and heterophilic graphs, introduce the single-pass graph contrastive learning loss based on the property, and provide performance guarantees of the minimizer of the loss on downstream tasks. As a direct consequence of our analysis, we implement the Single-Pass Graph Contrastive Learning method (SP-GCL). Empirically, on 14 benchmark datasets with varying degrees of heterophily, the features learned by the SP-GCL can match or outperform existing strong baselines with significantly less computational overhead, which verifies the usefulness of our findings in real-world cases.