LocMoE+: Enhanced Router with Token Feature Awareness for Efficient LLM Pre-Training

Mixture-of-Experts (MoE) architectures have recently gained increasing popularity within the domain of large language models (LLMs) due to their ability to significantly reduce training and inference overhead. However, MoE architectures face challenges, such as significant disparities in the number of tokens assigned to each expert and a tendency toward homogenization among experts, which adversely affects the model's semantic generation capabilities. In this paper, we introduce LocMoE+, a refined version of the low-overhead LocMoE, incorporating the following enhancements: (1) Quantification and definition of the affinity between experts and tokens. (2) Implementation of a global-level adaptive routing strategy to rearrange tokens based on their affinity scores. (3) Reestimation of the lower bound for expert capacity, which has been shown to progressively decrease as the token feature distribution evolves. Experimental results demonstrate that, without compromising model convergence or efficacy, the number of tokens each expert processes can be reduced by over 60%. Combined with communication optimizations, this leads to an average improvement in training efficiency ranging from 5.4% to 46.6%. After fine-tuning, LocMoE+ exhibits a performance improvement of 9.7% to 14.1% across the GDAD, C-Eval, and TeleQnA datasets.

EFSA: Towards Event-Level Financial Sentiment Analysis

In this paper, we extend financial sentiment analysis~(FSA) to event-level since events usually serve as the subject of the sentiment in financial text. Though extracting events from the financial text may be conducive to accurate sentiment predictions, it has specialized challenges due to the lengthy and discontinuity of events in a financial text. To this end, we reconceptualize the event extraction as a classification task by designing a categorization comprising coarse-grained and fine-grained event categories. Under this setting, we formulate the \textbf{E}vent-Level \textbf{F}inancial \textbf{S}entiment \textbf{A}nalysis~(\textbf{EFSA} for short) task that outputs quintuples consisting of (company, industry, coarse-grained event, fine-grained event, sentiment) from financial text. A large-scale Chinese dataset containing $12,160$ news articles and $13,725$ quintuples is publicized as a brand new testbed for our task. A four-hop Chain-of-Thought LLM-based approach is devised for this task. Systematically investigations are conducted on our dataset, and the empirical results demonstrate the benchmarking scores of existing methods and our proposed method can reach the current state-of-the-art. Our dataset and framework implementation are available at https://anonymous.4open.science/r/EFSA-645E

Transform then Explore: a Simple and Effective Technique for Exploratory Combinatorial Optimization with Reinforcement Learning

Many complex problems encountered in both production and daily life can be conceptualized as combinatorial optimization problems (COPs) over graphs. Recent years, reinforcement learning (RL) based models have emerged as a promising direction, which treat the COPs solving as a heuristic learning problem. However, current finite-horizon-MDP based RL models have inherent limitations. They are not allowed to explore adquately for improving solutions at test time, which may be necessary given the complexity of NP-hard optimization tasks. Some recent attempts solve this issue by focusing on reward design and state feature engineering, which are tedious and ad-hoc. In this work, we instead propose a much simpler but more effective technique, named gauge transformation (GT). The technique is originated from physics, but is very effective in enabling RL agents to explore to continuously improve the solutions during test. Morever, GT is very simple, which can be implemented with less than 10 lines of Python codes, and can be applied to a vast majority of RL models. Experimentally, we show that traditional RL models with GT technique produce the state-of-the-art performances on the MaxCut problem. Furthermore, since GT is independent of any RL models, it can be seamlessly integrated into various RL frameworks, paving the way of these models for more effective explorations in the solving of general COPs.

Satellite Federated Edge Learning: Architecture Design and Convergence Analysis

The proliferation of low-earth-orbit (LEO) satellite networks leads to the generation of vast volumes of remote sensing data which is traditionally transferred to the ground server for centralized processing, raising privacy and bandwidth concerns. Federated edge learning (FEEL), as a distributed machine learning approach, has the potential to address these challenges by sharing only model parameters instead of raw data. Although promising, the dynamics of LEO networks, characterized by the high mobility of satellites and short ground-to-satellite link (GSL) duration, pose unique challenges for FEEL. Notably, frequent model transmission between the satellites and ground incurs prolonged waiting time and large transmission latency. This paper introduces a novel FEEL algorithm, named FEDMEGA, tailored to LEO mega-constellation networks. By integrating inter-satellite links (ISL) for intra-orbit model aggregation, the proposed algorithm significantly reduces the usage of low data rate and intermittent GSL. Our proposed method includes a ring all-reduce based intra-orbit aggregation mechanism, coupled with a network flow-based transmission scheme for global model aggregation, which enhances transmission efficiency. Theoretical convergence analysis is provided to characterize the algorithm performance. Extensive simulations show that our FEDMEGA algorithm outperforms existing satellite FEEL algorithms, exhibiting an approximate 30% improvement in convergence rate.

LocMoE: A Low-overhead MoE for Large Language Model Training

The Mixtures-of-Experts (MoE) model is a widespread distributed and integrated learning method for large language models (LLM), which is favored due to its ability to sparsify and expand models efficiently. However, the performance of MoE is limited by load imbalance and high latency of All-To-All communication, along with relatively redundant computation owing to large expert capacity. Load imbalance may result from existing routing policies that consistently tend to select certain experts. The frequent inter-node communication in the All-To-All procedure also significantly prolongs the training time. To alleviate the above performance problems, we propose a novel routing strategy that combines load balance and locality by converting partial inter-node communication to that of intra-node. Notably, we elucidate that there is a minimum threshold for expert capacity, calculated through the maximal angular deviation between the gating weights of the experts and the assigned tokens. We port these modifications on the PanGu-Sigma model based on the MindSpore framework with multi-level routing and conduct experiments on Ascend clusters. The experiment results demonstrate that the proposed LocMoE reduces training time per epoch by 12.68% to 22.24% compared to classical routers, such as hash router and switch router, without impacting the model accuracy.

Comprehensive evaluation of deep and graph learning on drug-drug interactions prediction

Recent advances and achievements of artificial intelligence (AI) as well as deep and graph learning models have established their usefulness in biomedical applications, especially in drug-drug interactions (DDIs). DDIs refer to a change in the effect of one drug to the presence of another drug in the human body, which plays an essential role in drug discovery and clinical research. DDIs prediction through traditional clinical trials and experiments is an expensive and time-consuming process. To correctly apply the advanced AI and deep learning, the developer and user meet various challenges such as the availability and encoding of data resources, and the design of computational methods. This review summarizes chemical structure based, network based, NLP based and hybrid methods, providing an updated and accessible guide to the broad researchers and development community with different domain knowledge. We introduce widely-used molecular representation and describe the theoretical frameworks of graph neural network models for representing molecular structures. We present the advantages and disadvantages of deep and graph learning methods by performing comparative experiments. We discuss the potential technical challenges and highlight future directions of deep and graph learning models for accelerating DDIs prediction.

Bounded KRnet and its applications to density estimation and approximation

In this paper, we develop an invertible mapping, called B-KRnet, on a bounded domain and apply it to density estimation/approximation for data or the solutions of PDEs such as the Fokker-Planck equation and the Keller-Segel equation. Similar to KRnet, the structure of B-KRnet adapts the triangular form of the Knothe-Rosenblatt rearrangement into a normalizing flow model. The main difference between B-KRnet and KRnet is that B-KRnet is defined on a hypercube while KRnet is defined on the whole space, in other words, we introduce a new mechanism in B-KRnet to maintain the exact invertibility. Using B-KRnet as a transport map, we obtain an explicit probability density function (PDF) model that corresponds to the pushforward of a prior (uniform) distribution on the hypercube. To approximate PDFs defined on a bounded computational domain, B-KRnet is more effective than KRnet. By coupling KRnet and B-KRnet, we can also define a deep generative model on a high-dimensional domain where some dimensions are bounded and other dimensions are unbounded. A typical case is the solution of the stationary kinetic Fokker-Planck equation, which is a PDF of position and momentum. Based on B-KRnet, we develop an adaptive learning approach to approximate partial differential equations whose solutions are PDFs or can be regarded as a PDF. In addition, we apply B-KRnet to density estimation when only data are available. A variety of numerical experiments is presented to demonstrate the effectiveness of B-KRnet.

Gradient-enhanced deep neural network approximations

We propose in this work the gradient-enhanced deep neural networks (DNNs) approach for function approximations and uncertainty quantification. More precisely, the proposed approach adopts both the function evaluations and the associated gradient information to yield enhanced approximation accuracy. In particular, the gradient information is included as a regularization term in the gradient-enhanced DNNs approach, for which we present similar posterior estimates (by the two-layer neural networks) as those in the path-norm regularized DNNs approximations. We also discuss the application of this approach to gradient-enhanced uncertainty quantification, and present several numerical experiments to show that the proposed approach can outperform the traditional DNNs approach in many cases of interests.

Adaptive deep density approximation for fractional Fokker-Planck equations

In this work, we propose adaptive deep learning approaches based on normalizing flows for solving fractional Fokker-Planck equations (FPEs). The solution of a FPE is a probability density function (PDF). Traditional mesh-based methods are ineffective because of the unbounded computation domain, a large number of dimensions and the nonlocal fractional operator. To this end, we represent the solution with an explicit PDF model induced by a flow-based deep generative model, simplified KRnet, which constructs a transport map from a simple distribution to the target distribution. We consider two methods to approximate the fractional Laplacian. One method is the Monte Carlo approximation. The other method is to construct an auxiliary model with Gaussian radial basis functions (GRBFs) to approximate the solution such that we may take advantage of the fact that the fractional Laplacian of a Gaussian is known analytically. Based on these two different ways for the approximation of the fractional Laplacian, we propose two models, MCNF and GRBFNF, to approximate stationary FPEs and MCTNF to approximate time-dependent FPEs. To further improve the accuracy, we refine the training set and the approximate solution alternately. A variety of numerical examples is presented to demonstrate the effectiveness of our adaptive deep density approaches.

Solving time dependent Fokker-Planck equations via temporal normalizing flow

In this work, we propose an adaptive learning approach based on temporal normalizing flows for solving time-dependent Fokker-Planck (TFP) equations. It is well known that solutions of such equations are probability density functions, and thus our approach relies on modelling the target solutions with the temporal normalizing flows. The temporal normalizing flow is then trained based on the TFP loss function, without requiring any labeled data. Being a machine learning scheme, the proposed approach is mesh-free and can be easily applied to high dimensional problems. We present a variety of test problems to show the effectiveness of the learning approach.