The paper presents a multi-power law for predicting loss curves across different learning rate schedules.
problem Understanding and optimizing the relationship between model performance and hyperparameters, especially learning rates.
method Proposes a multi-power law that combines power laws based on the sum of learning rates and additional laws for loss reduction due to decay.
result The multi-power law accurately predicts loss curves for unseen learning rate schedules and finds a schedule that outperforms cosine learning rate.
LLMs learn peaked distributions slowly due to power-law losses.
problem Slow convergence of loss in training large language models.
method Systematic analysis of toy models and empirical evaluation of LLMs.
result Power-law time scaling with an exponent of 1/3 for learning peaked distributions.
Power of network tests degrades when vertices are misaligned.
problem Power loss in network hypothesis testing due to vertex shuffling.
method Theoretical analysis and simulations of Frobenius norm differences in random dot product and stochastic block models.
result Shuffling vertices can significantly reduce the power of network tests.
We develop a novel method, called PoWER-BERT, for improving the inference time of the popular BERT model, while maintaining the accuracy. It works by: a) exploiting redundancy pertaining to word-vectors (intermediate encoder outputs) and eliminating the redundant vectors. b) determining which word-vectors to eliminate …
New findings show neural network training loss follows a power law over time.
problem Understanding the optimization process of neural networks during training.
method Spectral analysis of the integral operator representing the linearized evolution of a large network.
result The loss function in neural network training follows a power law behavior, L(t)∼t−ξ, with exponent ξ determined by network parameters and data characteristics. Novel power transform unifies various mathematical functions.
problem Normalizing and standardizing datasets.
method Presented a novel power transform.
result Unified various mathematical functions.
Electronic power inverters are capable of quickly delivering reactive power to maintain customer voltages within operating tolerances and to reduce system losses in distribution grids. This paper proposes a systematic and data-driven approach to determine reactive power inverter output as a function of local measuremen…
Explains deep learning models and their geometric properties.
problem Understanding the geometric intuition behind deep learning models.
method Geometrical intuition and novel insights into loss surfaces of deep learning models.
result Deep neural networks carve out manifolds with multiplication neurons.
The paper predicts loss scaling across different datasets and compute scales.
problem Predicting loss scaling across different datasets and compute scales.
method Derive shifted power law relationships between train and test losses.
result Shifted power law relationships hold for various datasets and tasks, improving prediction accuracy.
Boosts change-point detection power with optimal sub-sampling.
problem Power loss in sequential change-point detection from large history data.
method Optimal sub-sampling of history data before kernel-based detection procedures.
result Improved detection performance in extensive experiments.
Federated learning calibrates insurance indices from renewable energy producers' data.
problem Calibrating parametric insurance indices under heterogeneous renewable energy production losses.
method Federated learning framework using Tweedie GLMs and distributed optimization.
result Federated learning recovers comparable index coefficients under moderate heterogeneity.
The paper analyzes a private likelihood-ratio test for frequency tables under differential privacy constraints.
problem Achieving privacy in statistical data analysis while maintaining statistical utility.
method A rigorous analysis of a private likelihood-ratio (LR) test for goodness-of-fit in frequency tables, considering (ε,δ)-differential privacy. result Characterization of the trade-off between differential privacy parameters (ε,δ) and statistical power of the private LR test. Correntropy is a second order statistical measure in kernel space, which has been successfully applied in robust learning and signal processing. In this paper, we define a nonsecond order statistical measure in kernel space, called the kernel mean-p power error (KMPE), including the correntropic loss (CLoss) as a speci…
This paper reformulates systemic risk measures and finds new properties and estimators.
problem Understanding and measuring systemic risk in financial networks.
method Representation of systemic risk measures in terms of univariate risk measures and quantiles determined by copulas. Empirical properties and estimators derived.
result MES is not suitable for measuring extreme risks. ES-based measures are more sensitive to power-law tails and large losses.
We discover scaling laws for kernel regression loss under various learning rate schedules.
problem Understanding loss dynamics and learning rate schedules in kernel regression.
method Theoretical analysis of stochastic gradient descent on a power-law kernel regression model.
result Established a Functional Scaling Law (FSL) capturing the full loss trajectory under arbitrary learning rate schedules.
A new AMM design reduces impermanent loss and retains more liquidity.
problem Inefficiencies in conventional AMM designs lead to liquidity loss and user engagement issues in DEXs.
method Proposes a dual-mechanism framework: a power-law invariant BMM and dynamic rebate system.
result Reduces impermanent loss by 36% and retains 3.98x more liquidity during price volatility.
Ensemble techniques are powerful approaches that combine several weak learners to build a stronger one. As a meta-learning framework, ensemble techniques can easily be applied to many machine learning methods. Inspired by ensemble techniques, in this paper we propose an ensemble loss functions applied to a simple regre…
Deep neural networks (DNNs) have great expressive power, which can even memorize samples with wrong labels. It is vitally important to reiterate robustness and generalization in DNNs against label corruption. To this end, this paper studies the 0-1 loss, which has a monotonic relationship with an empirical adversary (r…
Estimation of the operational risk capital under the Loss Distribution Approach requires evaluation of aggregate (compound) loss distributions which is one of the classic problems in risk theory. Closed-form solutions are not available for the distributions typically used in operational risk. However with modern comput…
We consider the problem of option hedging in a market with proportional transaction costs. Since super-replication is very costly in such markets, we replace perfect hedging with an expected loss constraint. Asymptotic analysis for small transactions is used to obtain a tractable model. A general expansion theory is de…
Paper introduces new loss functions for Siamese networks using FDA.
problem Training Siamese networks with improved loss functions.
method Proposes Fisher Discriminant Triplet (FDT) and Fisher Discriminant Contrastive (FDC) loss functions based on FDA.
result Shows effectiveness of FDT and FDC on MNIST and histopathology datasets.
The one-bit quantization is implemented by one single comparator that operates at low power and a high rate. Hence one-bit compressive sensing (1bit-CS) becomes attractive in signal processing. When measurements are corrupted by noise during signal acquisition and transmission, 1bit-CS is usually modeled as minimizing …
A meticulous assessment of the risk of impacts associated with extreme wind events is of great necessity for populations, civil authorities as well as the insurance industry. Using the concept of spatial risk measure and related set of axioms introduced by Koch (2017, 2019), we quantify the risk of losses due to extrem…
The paper sets lower bounds for adversarial robustness in multiclass classification.
problem Adversarial robustness in multiclass classification with arbitrary loss functions.
method Dual and barycentric reformulations for robust risk minimization.
result Sharp lower bounds for adversarial risks are computed efficiently.
Domain adaptation provides a powerful set of model training techniques given domain-specific training data and supplemental data with unknown relevance. The techniques are useful when users need to develop models with data from varying sources, of varying quality, or from different time ranges. We build CrossTrainer, a…
This paper introduces Stochastic Gradient Langevin Boosting (SGLB) - a powerful and efficient machine learning framework that may deal with a wide range of loss functions and has provable generalization guarantees. The method is based on a special form of the Langevin diffusion equation specifically designed for gradie…
Value functions struggle to represent transition dynamics, impacting statistical efficiency.
problem Limited representational power of value functions in capturing transition dynamics.
method Case studies of various reinforcement learning problems to explore the limitations of value-based methods.
result Value-based methods can be as efficient as model-based ones in some cases but severely underperform in others due to information loss.
Operational risk is the risk relative to monetary losses caused by failures of bank internal processes due to heterogeneous causes. A dynamical model including both spontaneous generation of losses and generation via interactions between different processes is presented; the efforts made by the bank to avoid the occurr…
We stabilize MI-based losses by adding a regularization term, improving their performance and stability.
problem Instability of MI-based losses in machine learning.
method Added a novel regularization term to stabilize MI-based losses.
result Regularization stabilizes training and improves the performance of MI-based losses.
Neural network predicts daily power consumption with high accuracy.
problem Middle-term power consumption prediction in the energy sector.
method Incorporates trend, seasonality, and weather conditions in a shallow Neural Network.
result Excellent density forecast results on one-year test set.
The classical asymptotic theory for parametric M-estimators guarantees that, in the limit of infinite sample size, the excess risk has a chi-square type distribution, even in the misspecified case. We demonstrate how self-concordance of the loss allows to characterize the critical sample size sufficient to guarantee …
Anonymization reduces economic signal extraction from financial texts.
problem Reducing meaningful economic signals from financial texts due to anonymization.
method Analyzed the impact of anonymization on textual understanding and economic signal extraction.
result Information loss due to anonymization is severe and pervasive, outweighing its benefits in certain financial applications.
We propose a dynamical model for the estimation of Operational Risk in banking institutions. Operational Risk is the risk that a financial loss occurs as the result of failed processes. Examples of operational losses are the ones generated by internal frauds, human errors or failed transactions. In order to encompass t…
This work improves diffusion models by estimating the optimal loss value for better training diagnostics.
problem The optimal loss value of diffusion models is unknown and not indicative of absolute data-fitting quality.
method Derive the optimal loss in closed form and develop effective estimators, including a stochastic variant.
result Unlocking the optimal loss as a metric for diagnosing training quality of diffusion models.
Paper explores challenges in training PINNs and loss landscape effects.
problem Challenges in training Physics-Informed Neural Networks (PINNs) due to loss landscape issues.
method Examined gradient-based optimizers Adam, L-BFGS, and their combination Adam+L-BFGS, and introduced NysNewton-CG (NNCG).
result Adam+L-BFGS outperforms other optimizers, and NysNewton-CG significantly improves PINN performance.
Loss-calibrated EP improves Bayesian decision-making by focusing on utility-sensitive posterior approximations.
problem Bayesian decision-making under asymmetric utility functions.
method Loss-calibrated expectation propagation (Loss-EP) that tilts the posterior towards higher utility decisions.
result Loss-EP can capture useful information for decision-making under asymmetric penalties.
We study the problem of nonparametric dependence detection. Many existing methods may suffer severe power loss due to non-uniform consistency, which we illustrate with a paradox. To avoid such power loss, we approach the nonparametric test of independence through the new framework of binary expansion statistics (BEStat…
Optimal exit strategies of CPT gamblers in unfair gambles
problem Optimal exit strategies of gamblers with CPT preferences in games with strictly negative expected payoffs
method Formulating the problem as an optimal stopping problem on asymmetric random walks, applying geometric transformation, randomized strategies, and changing the decision variable
result The unfair problem in the infinite time horizon has finite values for a wide range of CPT parameter specifications
Method detects and locates eavesdropping in optical links.
problem Detect and locate eavesdropping in optical links with small power losses.
method Cluster-based approach using OPM data at receiver and in-line OPM data for localization.
result Subtle eavesdropping losses can be detected and localized using OPM data.
Generative Adversarial Networks (GANs) were intuitively and attractively explained under the perspective of game theory, wherein two involving parties are a discriminator and a generator. In this game, the task of the discriminator is to discriminate the real and generated (i.e., fake) data, whilst the task of the gene…
SGD's escape rate depends on log loss barrier, not linear loss barrier.
problem Understanding the escape rate of SGD from local minima.
method Derived a stochastic differential equation (SDE) with additive noise from SGD's multiplicative noise property.
result The log loss barrier determines the escape rate of SGD, not the linear loss barrier.
We present a powerful new loss function and training scheme for learning binary hash codes with any differentiable model and similarity function. Our loss function improves over prior methods by using log likelihood loss on top of an accurate approximation for the probability that two inputs fall within a Hamming dista…
Study compares deep learning stock trading strategies in adverse market conditions.
problem Comparing deep learning models for stock trading performance in extreme market downturns.
method Reconstructed three deep learning models and compared their strategies through trading simulations.
result Deep learning models, especially LSTM, can mitigate losses in severe market downturns.
Ensemble techniques are powerful approaches that combine several weak learners to build a stronger one. As a meta learning framework, ensemble techniques can easily be applied to many machine learning techniques. In this paper we propose a neural network extended with an ensemble loss function for text classification. …
This paper develops a novel graph convolutional network (GCN) framework for fault location in power distribution networks. The proposed approach integrates multiple measurements at different buses while taking system topology into account. The effectiveness of the GCN model is corroborated by the IEEE 123 bus benchmark…
Locational Marginal Pricing aims to free UK power markets.
problem Unfree and regulated power markets.
method Implementing Locational Marginal Pricing.
result Increased economic freedom, reduced prices, decreased losses, incentivized investment.
We propose a reinforcement learning (RL) based closed loop power control algorithm for the downlink of the voice over LTE (VoLTE) radio bearer for an indoor environment served by small cells. The main contributions of our paper are to 1) use RL to solve performance tuning problems in an indoor cellular network for voic…
OptCS optimizes model selection after conformal inference, controlling FDR and power loss.
problem Challenges in model selection for conformal inference, especially when limited labeled data and many model choices are available.
method OptCS framework that allows valid statistical testing after flexible data-driven model optimization, using novel multiple testing procedures.
result Valid conformal p-values constructed despite substantial data reuse, maintaining FDR control.