Improved HMoE models using Laplace gating function enhance expert specialization and performance.
problem Improving performance of hierarchical mixture of experts models.
method Used Laplace gating function instead of Softmax in hierarchical mixture of experts models.
result Laplace gating function accelerates expert convergence and enhances specialization.
DriftMoE adapts to concept drifts in data streams efficiently.
problem Adapting to non-stationary data streams with concept drifts.
method Online Mixture-of-Experts (MoE) architecture with co-training framework.
result DriftMoE achieves competitive results with state-of-the-art stream learning adaptive ensembles.
The goal of meta-learning is to train a model on a variety of learning tasks, such that it can adapt to new problems within only a few iterations. Here we propose a principled information-theoretic model that optimally partitions the underlying problem space such that specialized expert decision-makers solve the result…
We develop the setting of sequential prediction based on shifting experts and on a "smooth" version of the method of specialized experts. To aggregate experts predictions, we use the AdaHedge algorithm, which is a version of the Hedge algorithm with adaptive learning rate, and extend it by the meta-algorithm Fixed Shar…
A new model improves the performance of knowledge distillation in GNNs.
problem Inconsistent performance of existing knowledge distillation techniques in GNNs.
method Proposes a new model, Routing-by-Memory (RbM), a form of Mixture-of-Experts (MoE), to address performance concerns.
result Demonstrates experimentally that RbM achieves considerably more consistent performance across multiple datasets.
This study analyzes communication constraints in MoE architectures using information theory.
problem Communication constraints in Mixture-of-Experts (MoE) architectures.
method Developed a rate-distortion characterization of finite-rate gating in MoE architectures using information theory.
result Yielded capacity-aware limits for communication-constrained MoE systems.
Specialized hierarchies improve decision-making by partitioning tasks.
problem Division of labor and specialization in hierarchical systems.
method Information-theoretic on-line learning rule for partitioning tasks.
result Enforces specialization and optimal combination of experts.
fMRI semantic category understanding using linguistic encoding models attempts to learn a forward mapping that relates stimuli to the corresponding brain activation. State-of-the-art encoding models use a single global model (linear or non-linear) to predict brain activation given the stimulus. However, the critical as…
A method to improve LLMs by automating the construction of a mixture of expert prompts.
problem Limitation of single instruction prompts in covering complex problem spaces.
method Divide the problem space into sub-regions, each governed by a specialized expert with both an instruction and demos. A two-phase process constructs these experts.
result Achieves an average win rate of 81% across major benchmarks.
Improved regret bounds for bandits with fixed expert advice using information theory.
problem Optimizing regret in bandit problems with fixed expert distributions.
method Information-theoretic analysis and KL-divergence measures.
result First regret bounds for EXP4 that can get arbitrarily close to zero under certain conditions.
Improved time series forecasting with expert loss integration.
problem Enhancing time series forecasting accuracy and efficiency.
method Adaptive Mixture-of-Experts framework with expert-specific loss integration and online learning.
result Significantly improved forecasting accuracy and computational efficiency.
The paper analyzes convergence rates of softmax gating in MoE models.
problem The effectiveness and scalability of machine learning models using MoE.
method Convergence analysis of parameter and expert estimation under MoE with softmax gating and its variants.
result Theoretical results show polynomially many data points are needed for strong identifiability conditions, while exponential points are required for linear experts.
Unified framework for deferring queries to top-k experts, improving accuracy-cost trade-offs.
problem Limitation of existing L2D frameworks to single-expert deferral.
method Top-k Learning-to-Defer framework, including adaptive Top-k(x) variant. result Superior accuracy-cost trade-offs with multi-expert deferral.
New method combines FMEA and Bayesian Network for root cause analysis in lithium-ion battery production.
problem Complex cause-effect relationships in lithium-ion battery production.
method Combining FMEA with Bayesian Network to detect and resolve inconsistencies.
result Holistic method builds large-scale cross-process Bayesian Failure Network for root cause analysis.
Sharp bounds found on expert error in binary advice aggregation.
problem Aggregating binary advice from conditionally independent experts.
method Sharp upper and lower bounds on optimal error probability in asymmetric case.
result Sharp bounds recover and sharpen known results in symmetric case.
Moirai-MoE improves time series forecasting by automatically specializing tokens without human-defined frequency.
problem Unified training on time series data remains challenging due to heterogeneity and non-stationarity.
method Uses sparse mixture of experts (MoE) within Transformers to automatically specialize tokens for diverse time series patterns.
result Moirai-MoE outperforms existing foundation models in both in-distribution and zero-shot scenarios.
Optimal investment strategy with expert opinions in uncertain conditions.
problem Optimizing wealth in a model with unobservable drift and costly expert opinions.
method Embedding into a full information problem, using viscosity solutions and stochastic Perron's method.
result Constructing optimal trading and expert opinion strategies under sufficient regularity conditions.
Optimized deferral improves accuracy in imbalanced settings.
problem Imbalance in expert predictions leads to suboptimal performance in two-stage learning to defer.
method Developed novel cost-sensitive learning algorithms and margin-based loss functions tailored for expert imbalance.
result MILD algorithm shows clear improvements over baselines in image classification and LLM routing tasks.
Bayesian model fuses multiple classifiers with explicit correlation modeling.
problem Combining outputs of multiple classifiers with explicit correlation.
method Hierarchical Bayesian model with correlated Dirichlet distribution.
result Fused classifier performance can be Bayes optimal even for highly correlated base classifiers.
The paper analyzes EM for Mixtures of Experts and shows its equivalence to projected Mirror Descent.
problem Training Mixtures of Experts (MoE) models.
method Rigorously analyzes Expectation Maximization (EM) for MoE models using a Mirror Descent perspective.
result Derives new convergence results and identifies conditions for local linear convergence.
MM-DREX adapts LLM experts for financial trading via dynamic routing.
problem Challenges of non-stationary financial markets and static expert designs.
method MM-DREX uses a VLM-powered dynamic router to allocate expert weights and designs heterogeneous trading experts.
result Significantly outperforms 15 baselines across key metrics.
Unsupervised deep learning is one of the most powerful representation learning techniques. Restricted Boltzman machine, sparse coding, regularized auto-encoders, and convolutional neural networks are pioneering building blocks of deep learning. In this paper, we propose a new building block -- distributed random models…
A novel framework for regression with multiple experts, addressing challenges in infinite and continuous label spaces.
problem Challenges in regression with multiple experts due to the infinite and continuous nature of the label space.
method Introduces a novel framework for regression with deferral, analyzing both single-stage and two-stage scenarios with new surrogate loss functions.
result Proves H-consistency bounds for both single-stage and two-stage methods, providing stronger guarantees than Bayes consistency. We consider the setting of sequential prediction of arbitrary sequences based on specialized experts. We first provide a review of the relevant literature and present two theoretical contributions: a general analysis of the specialist aggregation rule of Freund et al. (1997) and an adaptation of fixed-share rules of He…
Proposes a Varying-Coefficient MoE model for analyzing dynamic data.
problem Inadequate constant coefficients in MoE models for dynamic settings.
method Varying-Coefficient Mixture of Experts (VCMoE) model with varying coefficients in gating and expert models.
result Established identifiability and consistency of the VCMoE model.
TradeExpert uses a mix of LLMs to predict stock movements.
problem Synthesizing insights from diverse financial data sources.
method A mix of four specialized LLMs analyzing different data types, with a General Expert LLM synthesizing the insights.
result TradeExpert outperforms existing benchmarks in stock movement prediction.
Adversarial MoE learns category-specific models for product search.
problem Variations in product features and importance across categories.
method Mixture of Experts with adversarial regularization and soft gating constraints.
result Improved clustering of gate output vectors and shared experts among similar categories.
Proposes method for eliciting non-parametric joint priors using normalizing flows.
problem Learning complex non-parametric joint priors for model parameters.
method Expert elicitation combined with normalizing flows for generative modeling.
result Framework supports elicitation of both parametric and non-parametric priors.
Thompson Sampling, one of the oldest heuristics for solving multi-armed bandits, has recently been shown to demonstrate state-of-the-art performance. The empirical success has led to great interests in theoretical understanding of this heuristic. In this paper, we approach this problem in a way very different from exis…
This thesis tackles learning with multi-class abstention and multi-expert deferral, improving model reliability and efficiency.
problem Improving model reliability and efficiency in large language models (LLMs) by leveraging multiple experts.
method Developed new surrogate losses and consistency guarantees for multi-class classification and regression with deferral.
result Strong consistency guarantees for surrogate losses in multi-class classification and regression with deferral.
Improves reward bounds for prediction with expert advice using abstention.
problem Prediction with expert advice under bandit feedback with abstention.
method CBA algorithm exploiting abstention to improve reward bounds.
result Achieved significant improvement in reward bounds for general confidence-rated predictors.
MoE outperforms vanilla networks in detecting latent cluster structures in regression tasks.
problem Detecting latent cluster structures in regression tasks with gradient-based learning.
method Theoretical analysis of Mixture of Experts (MoE) using stochastic gradient descent (SGD).
result MoE can successfully divide the problem into subproblems, allowing each expert to recover simpler functions corresponding to individual clusters.
Study improves privacy-preserving online prediction from experts with speed-ups.
problem Privacy-preserving online prediction from experts with speed-ups.
method Differentially private federated online prediction algorithms.
result Achieves m-fold regret speed-up with low-loss expert in federated setting. EQD model improves domain-specific QA by 0.6% to 10.5%.
problem Challenges in domain-specific quantitative reasoning for LLMs.
method Two-step fine-tuning framework guided by a reward function.
result EQD outperforms state-of-the-art models and prompting strategies.
PCGS-TF uses a Transformer to adaptively control expert switching in non-stationary environments.
problem Static regret is insufficient for strictly online prediction in non-stationary settings.
method Policy-Controlled Generalized Share (PCGS) with a Transformer as an update controller.
result PCGS-TF achieves the lowest dynamic regret in non-stationary families and expert pools.
We improve MoE models for classification with rigorous guarantees and practical methods.
problem Limited guarantees for stable maximum-likelihood training and model selection in softmax-gated MoE models.
method Derived a batch MM algorithm with closed-form updates, proved finite-sample rates, and developed a dendrogram selector.
result Achieved near-parametric optimal rates for parameter recovery and improved accuracy over baselines.
Hydra boosts efficiency for long-context reasoning in resource-constrained settings.
problem Quadratic complexity of transformers limits long-context reasoning in resource-constrained systems.
method Hydra uses a modular architecture with adaptive routing between sparse global attention, mixture-of-experts, and dual memories.
result Hydra achieves significant throughput and accuracy improvements for long-context reasoning.
Machine learning (ML) has become a vital part in many aspects of our daily life. However, building well performing machine learning applications requires highly specialized data scientists and domain experts. Automated machine learning (AutoML) aims to reduce the demand for data scientists by enabling domain experts to…
STAR improves equivariant and invariant representation learning by routing projection heads.
problem Redundant feature learning in equivariant and invariant representation learning.
method Soft Task-Aware Routing (STAR) for projection heads specialization.
result Lower canonical correlations between invariant and equivariant embeddings.
FinHEAR combines LLMs with human expertise for better financial decision-making.
problem Challenges in financial decision-making for language models.
method Multi-agent framework with specialized LLMs for historical analysis, event interpretation, and expert retrieval.
result FinHEAR outperforms baselines in financial tasks with higher accuracy and risk-adjusted returns.
New method uses temperature to control sparse MoE convergence rates.
problem Sparse MoE convergence rates are slow due to temperature interactions.
method Proposes a novel activation gate to improve convergence rates.
result Improved convergence rates to polynomial rates via novel gate.
PRISM-VQ combines financial priors with vector quantization for better stock prediction.
problem Predicting cross-sectional stock returns is hard due to low signal-to-noise ratios and changing market conditions.
method Integrates expert priors, vector-quantized latent factors, and dynamic factor loadings.
result Consistent improvements in cross-sectional return prediction and portfolio performance.
Proof shows local convexity implies global convexity in special geometric spaces.
problem Proving convexity in CAT(0) cubed complexes from local convexity.
method Analyzes vertex link structures to determine convexity.
result Local combinatorial properties determine global convexity.
Despite significant advances in artificial intelligence (AI) for computer vision, its application in medical imaging has been limited by the burden and limits of expert-generated labels. We used images from optical coherence tomography angiography (OCTA), a relatively new imaging modality that measures perfusion of the…
Efficient algorithms for online learning with changing action sets, achieving no-approximate-regret guarantees.
problem Online learning with sleeping experts/bandits, where only a subset of actions are available each time.
method Developed computationally efficient algorithms providing no-approximate-regret guarantees for the general problem and better approximation ratios for special cases.
result Achieved no-approximate-regret guarantees for the general sleeping expert/bandit problems and better approximation ratios for specific cases.
Automated detection of MS lesions improves to 67% with 7T MRI.
problem Accurate detection of small, scarce cortical lesions in MS patients.
method 3D U-Net with brain tissue segmentation, supervised training on 7T MRI.
result 67% lesion detection rate with 42% false positives.
Unified sampling approach for Bayesian imaging problems.
problem Sampling from complex prior and posterior distributions in Bayesian imaging.
method Gaussian latent machine model for efficient prior and posterior sampling.
result Unified and generalized sampling algorithms for various imaging problems.
Multi-expert L2D underfits more severely, requiring new methods.
problem Underfitting in multi-expert L2D settings.
method PiCCE (Pick the Confident and Correct Expert), a surrogate-based method.
result PiCCE effectively reduces multi-expert L2D to a single-expert-like problem, resolving underfitting.