Survey of methods to train deep architectures without E2EBP.
problem Training deep architectures without end-to-end backpropagation.
method Modular training, weakly modular hybrids.
result Provably optimal alternatives to E2EBP can match or surpass E2EBP performance.
Noise-driven neural networks emerge modular structures, improving robustness and generalization.
problem Artificial neural networks struggle with modular solutions, leading to poor generalization and robustness.
method Inspired by brain's modular architecture, the study uses neural noise and nonlinear responses to drive the emergence of modular solutions.
result Noise-driven modularisation improves robustness and generalization in neural networks.
Training a Neural Network (NN) with lots of parameters or intricate architectures creates undesired phenomena that complicate the optimization process. To address this issue we propose a first modular approach to NN design, wherein the NN is decomposed into a control module and several functional modules, implementing …
NACs learn modular neural architectures without domain knowledge.
problem Jointly learn module configuration and execution without domain knowledge.
method Jointly trains two systems: module configuration and execution.
result Improves low-shot adaptation and OOD robustness.
Neural networks learn modular arithmetic but not all, extending known solutions to generalize.
problem Neural networks struggle with modular arithmetic, especially for polynomials.
method Developed analytical solutions for MLP networks to learn modular addition and multiplication, then combined these solutions to generalize on arbitrary modular polynomials.
result Neural networks can learn and generalize solutions to modular polynomials, supporting the hypothesis that some polynomials are learnable.
Modular RL modules solve complex 3D Sokoban tasks.
problem Solving complex, integrated tasks combining visual, physical, and abstract reasoning.
method Compose RL modules in a sense-plan-act hierarchy, using only model-free methods.
result Modular RL outperforms state-of-the-art monolithic RL on Mujoban.
The study examines if ReLU activation function is optimal for modularity in neural networks.
problem Finding the best activation function for modularity in neural networks.
method Comparing ReLU with other activation functions for modularity and performance.
result ReLU may not be the best choice for modularity, suggesting other functions could be more suitable.
RNNs solve modular addition tasks using low rank and sparse Fourier structures.
problem Solving modular addition tasks with recurrent neural networks.
method Identified low rank structures and sparse Fourier representations in RNN weights.
result RNNs robust to removing individual frequencies but degrade with more ablation.
SDQL uses modular deep Q networks to efficiently learn multi-stage optimal control tasks.
problem Training complex deep reinforcement learning models for multi-stage control tasks is inefficient and unstable.
method Stacked Deep Q Learning (SDQL) with modular Q networks and backward training.
result SDQL efficiently learns optimal control policies for multi-stage tasks with high-dimensional state and action spaces.
FinRL-X unifies trading components for AI and rule-based strategies.
problem Inconsistent between research and live deployment in trading platforms.
method Modular architecture integrating data processing, strategy construction, backtesting, and execution.
result Unified protocol supports AI and rule-based trading components without altering execution.
New theory maps neural network weights to optimize faster and scale.
problem Optimizing neural networks for speed and scalability.
method Constructing a duality map using layer-wise operator norms.
result Derived GPU-friendly algorithms for various layers.
ProMoD models human race drivers with probabilistic movement primitives and neural networks.
problem Challenging task of modeling human driver behavior due to variability and complexity.
method Modular framework with Probabilistic Movement Primitives, clothoids, and neural networks.
result Significant advantages in imitation accuracy and robustness compared to other algorithms.
Neural architecture search methods are able to find high performance deep learning architectures with minimal effort from an expert. However, current systems focus on specific use-cases (e.g. convolutional image classifiers and recurrent language models), making them unsuitable for general use-cases that an expert migh…
BNNs enhance reservoir computing by acting as generalization filters.
problem Understanding how BNNs integrate with reservoir computing.
method Optogenetics and calcium imaging to record BNNs, reservoir computing framework.
result BNNs improve reservoir computing performance through generalization.
Attention mechanism combines bottom-up and top-down signals in neural networks.
problem Combining robust perception with bottom-up and top-down signals.
method Attention mechanism over modulated recurrent neural networks.
result Bidirectional information flow leads to improved performance in various tasks.
Neuro-inspired recurrent neural network algorithms, such as echo state networks, are computationally lightweight and thereby map well onto untethered devices. The baseline echo state network algorithms are shown to be efficient in solving small-scale spatio-temporal problems. However, they underperform for complex task…
A framework for modular training of robust generative models.
problem Training large generative models is resource-intensive and requires heuristic tuning.
method Modular training using a gating mechanism and a minimax game to find a robust gate.
result The modular approach can theoretically outperform monolithic baselines and is scalable.
This paper introduces a new system for discovering patterns in morphogenetic systems using modular architecture and unsupervised learning.
problem Discovering novel patterns in morphogenetic systems is challenging and often relies on manual tuning.
method Introduces a hierarchical, modular architecture for unsupervised learning of diverse representations combined with goal exploration algorithms.
result The new system efficiently adapts diversity search towards user preferences with minimal feedback.
Improved recurrent neural networks learn long-term dependencies through multi-scale memory.
problem Capturing long-term dependencies in recurrent neural networks.
method Incremental training of a modular RNN architecture with multi-scale hidden states.
result Incremental training and multi-scale memory enhance RNNs' ability to learn long-term dependencies.
We propose a modular extension of backpropagation for the computation of block-diagonal approximations to various curvature matrices of the training objective (in particular, the Hessian, generalized Gauss-Newton, and positive-curvature Hessian). The approach reduces the otherwise tedious manual derivation of these mat…
Contemporary sensorimotor learning approaches typically start with an existing complex agent (e.g., a robotic arm), which they learn to control. In contrast, this paper investigates a modular co-evolution strategy: a collection of primitive agents learns to dynamically self-assemble into composite bodies while also lea…
ModSSC unifies semi-supervised classification for various data types.
problem Fragmented support for semi-supervised classification across different methods, settings, and data types.
method ModSSC is a modular Python framework that supports reproducible and controlled experimentation for semi-supervised classification on heterogeneous data.
result ModSSC enables systematic comparison of semi-supervised learning across various datasets and model backbones.
We explore efficient neural architecture search methods and show that a simple yet powerful evolutionary algorithm can discover new architectures with excellent performance. Our approach combines a novel hierarchical genetic representation scheme that imitates the modularized design pattern commonly adopted by human ex…
A core aspect of human intelligence is the ability to learn new tasks quickly and switch between them flexibly. Here, we describe a modular continual reinforcement learning paradigm inspired by these abilities. We first introduce a visual interaction environment that allows many types of tasks to be unified in a single…
We propose an actor-critic, model-free, and online Reinforcement Learning (RL) framework for continuous-state continuous-action Markov Decision Processes (MDPs) when the reward is highly sparse but encompasses a high-level temporal structure. We represent this temporal structure by a finite-state machine and construct …
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.
NOMU improves neural network uncertainty estimation.
problem Estimating model uncertainty for neural networks with limited data.
method Introduces NOMU, a two-sub-NN architecture with a designed loss function.
result NOMU outperforms state-of-the-art methods in regression and Bayesian optimization.
A new method recovers rewards from behavior policies using classification and regression.
problem Recovering meaningful rewards from observed behavior in reinforcement learning.
method GenPQR, a modular procedure that estimates behavior policy, evaluates soft Q-function, and recovers normalized reward using classification and regression.
result GenPQR matches or improves reward recovery compared to DeepPQR, while being simpler and more modular.
Recursive Feature Machines show grokking in modular arithmetic without neural networks.
problem Grokking in modular arithmetic tasks.
method Recursive Feature Machines (RFM) with Average Gradient Outer Product (AGOP).
result RFM and neural networks learn block-circulant features to solve modular arithmetic.
MOCA uses modular attention to estimate causal effects from complex data.
problem Estimating causal effects from observational data with complex, non-linear, and high-dimensional treatment and outcome mechanisms.
method MOCA is a transformer-based framework that separates treatment and outcome modeling through modular design and one-way attention mechanism, with cutting-feedback to prevent outcome influence on treatment representations.
result MOCA outperforms classical estimators and machine learning approaches across various simulated and real-world scenarios.
We focus on two supervised visual reasoning tasks whose labels encode a semantic relational rule between two or more objects in an image: the MNIST Parity task and the colorized Pentomino task. The objects in the images undergo random translation, scaling, rotation and coloring transformations. Thus these tasks involve…
There has been a rapid progress in the task of Visual Question Answering with improved model architectures. Unfortunately, these models are usually computationally intensive due to their sheer size which poses a serious challenge for deployment. We aim to tackle this issue for the specific task of Visual Question Answe…
Proposes efficient deep causal generative models for high-dimensional causal inference.
problem Inefficient training of deep generative models for high-dimensional data.
method Modular training of deep causal generative models using adversarial training and pre-trained models.
result First algorithm that provably samples from any identifiable causal query in the presence of latent confounders.
We present LumièreNet, a simple, modular, and completely deep-learning based architecture that synthesizes, high quality, full-pose headshot lecture videos from instructor's new audio narration of any length. Unlike prior works, LumièreNet is entirely composed of trainable neural network modules to learn mapping functi…
An artificial agent for financial risk and returns' prediction is built with a modular cognitive system comprised of interconnected recurrent neural networks, such that the agent learns to predict the financial returns, and learns to predict the squared deviation around these predicted returns. These two expectations a…
As deep learning applications continue to become more diverse, an interesting question arises: Can general problem solving arise from jointly learning several such diverse tasks? To approach this question, deep multi-task learning is extended in this paper to the setting where there is no obvious overlap between task a…
Convolutional architectures have recently been shown to be competitive on many sequence modelling tasks when compared to the de-facto standard of recurrent neural networks (RNNs), while providing computational and modeling advantages due to inherent parallelism. However, currently there remains a performance gap to mor…
CoE modularizes LLMs for scalable, cost-effective AI systems.
problem Challenges in scalability, cost, and customization of monolithic LLMs.
method Composition of Experts (CoE) leverages multiple expert LLMs with a router for dynamic selection.
result CoE achieves superior performance with reduced computational overhead and cost.
We find and propose an explanation for a large variety of modularity-related symmetries in problems of 3-manifold topology and physics of 3d N=2 theories where such structures a priori are not manifest. These modular structures include: mock modular forms, SL(2,Z) Weil representations, quantum mo…
Researchers found the global topology of the Eisenstein-Picard modular surface.
problem Understanding the global topology of the Eisenstein-Picard modular surface.
method Quotient space of the complex hyperbolic plane by the modular group.
result Determined the global topology of the Eisenstein-Picard modular surface as a 4-orbifold.
Study modular surfaces in Lorentz-Minkowski 3-space, classifying and analyzing their curvature and applications.
problem Understanding the curvature properties of modular surfaces in Lorentz-Minkowski space.
method Analyzing the sign of Gaussian and mean curvature, classifying surfaces, and applying to conformal field theories.
result Complete classification of zero Gaussian curvature modular surfaces and non-existence of non-planar maximal modular surfaces.
Study on NNs for forecasting time series with novel control variable combinations.
problem Forecast future time series with novel combinations of control variables.
method Modular NN architecture with inductive bias for independence of control variables.
result Modular NN architecture improves forecasting of dependent variables up to large horizons.
Modular neural networks generalize better with less data.
problem Theoretical and practical understanding of how modularity improves neural network generalization.
method Theoretical analysis of sample complexity, development of a novel learning rule.
result Modular networks require fewer samples to generalize compared to nonmodular networks, especially in high-dimensional tasks.
Study modular forms over Γ^0(2) and anomaly cancellation formulas.
problem Anomaly cancellation formulas for modular forms over Γ^0(2).
method Study and analysis of modular forms over Γ^0(2).
result Anomaly cancellation formulas derived for modular forms over Γ^0(2).
Our aim is to introduce and advocate non-Σ (non-symmetric) modular operads. While ordinary modular operads were inspired by the structure of the moduli space of stable complex curves, non-Σ modular operads model surfaces with open strings outputs. An immediate application of our theory is a short proof that the mod…
The common pipeline in autonomous driving systems is highly modular and includes a perception component which extracts lists of surrounding objects and passes these lists to a high-level decision component. In this case, leveraging the benefits of deep reinforcement learning for high-level decision making requires spec…
Fuchsian groups with a modular embedding have the richest arithmetic properties among non-arithmetic Fuchsian groups. But they are very rare, all known examples being related either to triangle groups or to Teichmueller curves. In Part I of this paper we study the arithmetic properties of the modular embedding and deve…
HiDe learns hierarchical control for complex tasks by separating planning and control.
problem Solving long horizon control tasks with generalization to unseen scenarios.
method Functional decomposition of state-action spaces, RL-based planner, modular transfer of policy layers.
result Generalizes across unseen test environments and scales to longer horizons.