Geometric insights reveal transferable adversarial directions across classifiers and models.
problem Adversarial perturbations transfer between different inputs, models, and architectures.
method Geometric analysis of linear classifiers and two-layer ReLU networks.
result Transferable adversarial directions exist for linear separators and ReLU networks with high probability.
Neural networks learn faster with correlated latent variables.
problem Efficiently learning from higher-order correlations in neural networks.
method Analytical derivation and simulations of two-layer neural networks.
result Correlations between latent variables speed up learning from higher-order correlations.
A framework visualizes embedding spaces of neural survival analysis models using anchor directions.
problem Visualizing complex embeddings in neural survival analysis models.
method Estimating anchor directions through clustering or user-supplied concepts, revealing relationships with raw inputs and survival times.
result Visualization strategies reveal how anchor directions relate to raw clinical features and survival time distributions.
Interneurons improve learning in neural networks by accelerating convergence.
problem Rapid adaptation to changing input statistics in neural networks.
method Two mathematically tractable recurrent linear neural networks were compared: one with direct recurrent connections and the other with interneurons that mediate recurrent communication.
result The network with interneurons converges more quickly than the network with direct recurrent connections, scaling logarithmically with initialization spectrum.
LayerNorm transformers have dead directions that can be read from their parameters alone.
problem Locating dead directions in LayerNorm transformers
method Using the inverse-scale direction of LayerNorm affine parameters
result Predicted dead direction matches measured bottom singular direction
Representation mixing combines character and phoneme inputs for flexible TTS synthesis.
problem Limited control over pronunciation in character or phoneme-based TTS systems.
method Representation mixing combines multiple linguistic inputs in a single encoder.
result Flexibility in choosing between character, phoneme, or mixed representations during inference.
Novel Bayesian neural network method for robustness.
problem Adversarial robustness without online training.
method Distributes uncertainty across all inputs.
result Demonstrates robustness on benchmark datasets.
From a machine learning perspective, the human ability localize sounds can be modeled as a non-parametric and non-linear regression problem between binaural spectral features of sound received at the ears (input) and their sound-source directions (output). The input features can be summarized in terms of the individual…
Unified routing and arbitrage with concave continuation.
problem Combining routing and arbitrage in financial markets.
method Extending AMM trade functions to negative inputs via concave continuation.
result Unified approach unifies routing and arbitrage.
An intriguing property of deep neural networks is their inherent vulnerability to adversarial inputs, which significantly hinders their application in security-critical domains. Most existing detection methods attempt to use carefully engineered patterns to distinguish adversarial inputs from their genuine counterparts…
Following great success in the image processing field, the idea of adversarial training has been applied to tasks in the natural language processing (NLP) field. One promising approach directly applies adversarial training developed in the image processing field to the input word embedding space instead of the discrete…
CASTLE learns causal DAG to improve model generalization.
problem Improving model generalization to out-of-sample data.
method CASTLE learns causal relationships via adjacency matrix embedded in neural network input layers, reconstructing only causal features.
result CASTLE leads to better out-of-sample predictions compared to other regularizers.
New algorithm reduces matrix multiplication time for sparse matrices.
problem Efficiently multiply large sparse matrices with limited space.
method Exploits sparsity to reduce QR decompositions and time complexity.
result Time complexity reduced to $\widetilde{O}\left((
nz(X)+
nz(Y))\ell+n\ell^2
ight)$ in expectation.
The contractive auto-encoder learns a representation of the input data that captures the local manifold structure around each data point, through the leading singular vectors of the Jacobian of the transformation from input to representation. The corresponding singular values specify how much local variation is plausib…
The Renormalisation Group (RG) provides a framework in which it is possible to assess whether a deep-learning network is sensitive to small changes in the input data and hence prone to error, or susceptible to adversarial attack. Distinct classification outputs are associated with different RG fixed points and sensitiv…
SpecGD mitigates misalignment in phase retrieval models with anisotropic inputs.
problem Misalignment during gradient descent in phase retrieval models with anisotropic inputs.
method Spectral gradient descent modifies gradient updates to preserve directional information and remove spike amplification.
result SpecGD removes spike amplification, leading to stable alignment and accelerated noise contraction.
The goal of this paper is to analyze the geometric properties of deep neural network classifiers in the input space. We specifically study the topology of classification regions created by deep networks, as well as their associated decision boundary. Through a systematic empirical investigation, we show that state-of-t…
Neural networks with DAGs show linearity as width increases.
problem Understanding linearity in neural networks with arbitrary DAG structures.
method Analyzing the transition to linearity in networks with arbitrary DAGs, characterizing width by minimum in-degree.
result General neural networks with DAGs exhibit linearity as width approaches infinity.
A multilayer perceptron can behave as a generative classifier by applying bidirectional learning (BL). It consists of training an undirected neural network to map input to output and vice-versa; therefore it can produce a classifier in one direction, and a generator in the opposite direction for the same data. The lear…
Paper proposes a novel data augmentation method for neural networks in DOA estimation.
problem Improving neural network performance in Direction of Arrival (DOA) estimation.
method Applies a FOA-based transformation to input data and labels to simulate new DOA information.
result Improves DOA estimation error by around 40%.
Quantum algorithm finds extremal values without direct function access.
problem Finding extremal values of hidden functions without direct access.
method Parametric quantum circuit trained with a trainable quantum feature map.
result Algorithm successfully finds extremal values even with sparse training data.
TQF models multivariate uncertainty by learning conditional quantiles.
problem Challenges in fully nonparametric estimation of multivariate conditional distributions.
method Tomographic Quantile Forests (TQF) learns conditional quantiles of directional projections.
result TQF reconstructs multivariate conditional distribution efficiently without convexity restrictions.
New framework interprets deep neural networks through input structure.
problem Limited understanding of how input structure, network parameters, and optimization algorithms work together.
method Introducing a novel theoretical framework based on the compositional structure of piecewise linear activation functions.
result Shows that input instances can be grouped based on their similarity in the internal representation of the neural network.
Machine learning models predict EUR/USD currency direction with 58.52% accuracy.
problem Predicting the directional movement of EUR/USD in the Foreign Exchange market.
method Comparative analysis of machine learning models, including decorrelated and non-decorrelated feature sets, and meta-estimators.
result 58.52% accuracy for one-day ahead forecasts.
Using artificial neural network for the prediction of heat demand has attracted more and more attention. Weather conditions, such as ambient temperature, wind speed and direct solar irradiance, have been identified as key input parameters. In order to further improve the model accuracy, it is of great importance to und…
The training of deep neural network classifiers results in decision boundaries which geometry is still not well understood. This is in direct relation with classification problems such as so called adversarial examples. We introduce zoNNscan, an index that is intended to inform on the boundary uncertainty (in terms of …
Study reveals how neural network architectures bias their learning based on feature directions.
problem Understanding how neural network architectures bias their learning based on feature directions.
method Defined neural anisotropy directions (NADs) to encapsulate the directional inductive bias of architectures and provided an efficient method to identify them.
result NADs characterize the features used by CNNs to discriminate between different classes for the CIFAR-10 dataset.
MTL2L learns to adapt optimisation rules for unseen data.
problem Learners need to adapt to unseen data domains.
method Introduces MTL2L, a context-aware neural optimiser.
result MTL2L can adapt optimisation rules for unseen data.
PGD-trained models have a preferential direction in their gradients, which improves robustness.
problem Mathematical lack of clarity in the direction of preferential gradient alignment after adversarial training.
method Proposed a novel definition of preferential direction and evaluated it using a metric based on GANs.
result PGD-trained models have higher alignment with the proposed preferential direction than baseline models.
The paper tackles stable maxima optimization for expensive functions.
problem Finding stable maxima of expensive functions with input variations.
method Uses multiple gradient Gaussian Process models to estimate stability and guide optimization.
result Demonstrates effective finding of stable maxima on synthetic and real-world problems.
Generates high-resolution images from low-resolution inputs.
problem Generating realistic images from low-resolution inputs.
method Latent Adversarial Generator (LAG) using perceptual loss.
result Samples of high-resolution images from low-resolution inputs.
Reconstructing polytopes with fixed facet directions from support function evaluations.
problem Reconstructing polytopes with known facet directions from limited data.
method Least-squares estimate via convex quadratic program, combinatorial characterization for uniqueness, algorithm convergence.
result The least-squares estimate for a fixed simplicial normal fan is a convex quadratic program, and the solution is unique under certain conditions.
Kernel models learn low-dimensional predictive subspaces from input data.
problem Learning effective feature transformations in kernel models.
method Study of a compositional kernel ridge regression model.
result Global minimizers of the objective function identify the subspace with high probability.
I-BERT extends Transformer's self-attention to arbitrary input lengths.
problem Transformer models struggle with inductive generalization to unseen input lengths.
method Replaces positional encodings with a recurrent layer.
result I-BERT achieves state-of-the-art results on algorithmic tasks.
FoRDE uses input gradients to improve neural network ensembles.
problem Improving neural network ensembles for robustness and accuracy.
method Proposes FoRDE, an ensemble learning method based on ParVI, which repels function space by input gradients.
result FoRDE significantly outperforms DEs and other ensemble methods in accuracy and calibration.
New Bayesian neural network model improves performance and compresses networks.
problem Improving predictive performance and compressing neural networks.
method Proposes a new variational family with radial and directional components.
result Improves predictive performance and yields compressed networks.
We present a novel algorithm (Principal Sensitivity Analysis; PSA) to analyze the knowledge of the classifier obtained from supervised machine learning techniques. In particular, we define principal sensitivity map (PSM) as the direction on the input space to which the trained classifier is most sensitive, and use anal…
This work investigates how gradient-based learning performs with structured data, revealing issues and improvements.
problem Gradient-based learning under structured data, particularly with a spiked covariance structure.
method Investigates the effect of a spiked covariance structure on gradient-based feature learning and proposes weight normalization.
result Gradient-based dynamics may fail to recover the true direction in anisotropic settings, but weight normalization can improve performance.
Trellis networks improve sequence modeling performance.
problem Sequence modeling challenges.
method Temporal convolutional network with weight tying and direct input injection.
result Trellis networks outperform state-of-the-art methods on benchmarks.
We develop a method for quantile-based sensitivity analysis in models with discontinuities.
problem Uncertainty in interpreting discontinuous models using traditional derivatives.
method Quantile-based derivatives for discontinuous models with discrete inputs.
result Derivatives of quantile-based outputs are well-defined and provide meaningful insights.
Proposes a framework to incorporate global sensitivity into local surrogate models.
problem Narrowing focus to local scale in surrogate modeling leads to re-learning global trends.
method Integrates global sensitivity analysis into local surrogate models through input warping.
result Local models become equally sensitive to all input directions, focusing on local dynamics.
Two impossibility theorems show formal alignment certification is impossible for AI systems.
problem Formal certification of AI alignment over open-ended domains is impossible.
method Two independent impossibility theorems: Semantic and Statistical barriers.
result No procedure can simultaneously satisfy soundness, completeness, and tractability.
Directed networks are pervasive both in nature and engineered systems, often underlying the complex behavior observed in biological systems, microblogs and social interactions over the web, as well as global financial markets. Since their structures are often unobservable, in order to facilitate network analytics, one …
We propose a new regularization method based on virtual adversarial loss: a new measure of local smoothness of the conditional label distribution given input. Virtual adversarial loss is defined as the robustness of the conditional label distribution around each input data point against local perturbation. Unlike adver…
This paper proposes a novel kernel approach to linear dimension reduction for supervised learning. The purpose of the dimension reduction is to find directions in the input space to explain the output as effectively as possible. The proposed method uses an estimator for the gradient of regression function, based on the…
Uses news sentiment scores for direct reinforcement trading in financial markets.
problem Incorporating news data into quantitative trading remains challenging.
method Directly uses news sentiment scores and raw data as inputs for reinforcement learning, processed by sequence models.
result Achieves superior performance compared to market benchmarks.
Adversarial examples are maliciously perturbed inputs designed to mislead machine learning (ML) models at test-time. They often transfer: the same adversarial example fools more than one model. In this work, we propose novel methods for estimating the previously unknown dimensionality of the space of adversarial inputs…
Fairness is a critical trait in decision making. As machine-learning models are increasingly being used in sensitive application domains (e.g. education and employment) for decision making, it is crucial that the decisions computed by such models are free of unintended bias. But how can we automatically validate the fa…