Research
On-device research index

arXiv research

A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.

169,051 papers · 148 categories

Trend · papers per month

2755508241,099 · Jun 202019922001200920172026
48 results for prototypical network loss

Prototype networks on hyperspheres improve classification and regression.

problem Improving classification and regression performance.
method Using hyperspherical prototypes for classification and regression, optimizing prototypes through data-independent margin separation.
result Hyperspherical prototype networks outperform other methods in classification, regression, and their combination.

Prototype model improves model auditing and understanding.

problem Auditing and understanding modern language models is expensive and approximate.
method Introduced a sparse, non-negative mixture of learned prototypes trained with clustering objectives.
result Prototype models either surpass or remain within 2.5 percentage points of dense baselines on downstream tasks.

HitNet uses a Hit-or-Miss layer to enhance feature interpretability in capsule networks.

problem Difficulty in interpreting complex neural network architectures.
method Replacing the last layer with a Hit-or-Miss layer that trains capsules to hit or miss a target capsule using centripetal loss.
result HitNet achieves better performance than initial CapsNet on various datasets and provides interpretable feature representations.

Prototype-based memory network learns visual categories from unlabeled data.

problem Learning from nonstationary, unlabeled data with sequential dependencies.
method Online prototype-based memory network with contrastive loss.
result Significantly better category recognition compared to state-of-the-art methods.

DDCL-INCRT learns its own structure during training, reducing unnecessary complexity.

problem Fixed-size neural network architectures require manual tuning, leading to overfitting.
method Combines DDCL (Deep Dual Competitive Learning) and INCRT (Incremental Transformer) to self-organize network structure.
result The network self-organizes into a hierarchy of heads, reducing unnecessary complexity.

We propose a novel architecture for kk-shot classification on the Omniglot dataset. Building on prototypical networks, we extend their architecture to what we call Gaussian prototypical networks. Prototypical networks learn a map between images and embedding vectors, and use their clustering for classification. In our…

2017-08-09abs ↗pdf ↗

A new neural network method improves interpretability and detection of outliers.

problem Improving interpretability and detection of outliers in neural networks.
method Prototype-based learning (PbL) using a winner-take-all (WTA) network with two prototypes: positive and negative.
result The negative prototype is similar to the positive one, aligning with the BCM theory.

We propose prototypical networks for the problem of few-shot classification, where a classifier must generalize to new classes not seen in the training set, given only a small number of examples of each new class. Prototypical networks learn a metric space in which classification can be performed by computing distances…

2017-03-15abs ↗pdf ↗

The key issue of few-shot learning is learning to generalize. This paper proposes a large margin principle to improve the generalization capacity of metric based methods for few-shot learning. To realize it, we develop a unified framework to learn a more discriminative metric space by augmenting the classification loss…

2018-07-08abs ↗pdf ↗

ProSeNet provides interpretable deep sequence models with natural explanations.

problem Challenges in explaining deep neural network predictions for sequence modeling.
method Prototypes derived from case-based reasoning, with criteria for simplicity, diversity, and sparsity.
result Achieves accuracy on par with state-of-the-art models while providing interpretable explanations.

Paper proposes integrating hierarchical class structure into prototypical network supervision.

problem Improving classification accuracy in tasks with hierarchical class structures.
method Integrates hierarchical class structure (metric) into prototypical network supervision.
result Consistent improvement of error rate weighted by the cost matrix compared to traditional methods.

ProtoryNet interprets text sequences using prototype trajectories for better understanding.

problem Improving text classification interpretability and accuracy.
method ProtoryNet uses prototype trajectories to interpret text sequences, with prototype pruning for better interpretability.
result ProtoryNet outperforms baseline models and reduces performance gap compared to black-box models.

Prototypical Networks improve multi-label classification accuracy.

problem Multi-label classification with nonlinear label dependencies.
method Formulate multi-label learning as class distribution in a non-linear embedding space. For each label, positive and negative embeddings are compactly distributed. Labels are inferred by measuring the distance to prototype positive or negative embeddings.
result Extensive experiments show improved accuracy compared to state-of-the-art algorithms.

Improved cover detection in music datasets with novel triplet loss.

problem Challenging task of automatically detecting covers in audio datasets.
method Convolutional neural network mapping melodic features to embeddings, training to minimize cover distance and maximize non-cover distance.
result New prototypical triplet loss improves accuracy for large datasets and live songs.

Geometric framework links clustering accuracy to structural recovery.

problem Understanding the trade-off between robustness and sensitivity in clustering.
method Develops a clustering condition number to compare within-cluster scale to the minimum loss increase required to move a point across a cluster boundary.
result Sharp phase transitions for exact recovery under different objectives, providing geometric principle for interpreting low objective values.

The paper analyzes the dynamics of a simple neural network using a mean-field approach.

problem Understanding the training dynamics of neural networks, especially in classification tasks.
method Developed an analytic theory using a mean-field limit for a simple neural network.
result Explicitly solved the dynamics of a linearly separable dataset with a linear hinge loss.

Advances few-shot classification by treating it as supervised learning and proposing new training techniques.

problem Formulating the ability of humans to learn from limited data in machine learning.
method Formulated few-shot classification as a supervised learning problem and introduced multi-episode and cross-way training techniques.
result Proposed training strategies accelerate the training process without accuracy loss.

Graph Prototypical Networks improve few-shot node classification on attributed networks.

problem Few-shot node classification in attributed networks with limited labeled instances.
method Graph Prototypical Networks (GPN) using meta-learning to extract meta-knowledge and identify informative labeled instances.
result GPN achieves superior performance in few-shot node classification.

ProtoNAM models tabular data with neural networks, making predictions transparent.

problem Tabular data analysis using neural networks lacks transparency and accuracy compared to tree-based methods.
method ProtoNAM introduces prototypes into neural networks to model tabular data while maintaining explainability.
result ProtoNAM outperforms existing NN-based GAMs and provides insights into learned feature patterns.

A new method improves few-shot learning by combining ProtoNet with LFD.

problem Few-shot learning struggles with high variance support sets.
method Combines ProtoNet with Local Fisher Discriminant Analysis.
result Superior classification accuracy on miniImageNet and tieredImageNet.

SLEEPER combines deep learning with expert rules for accurate sleep staging.

problem Manual sleep staging is tedious and requires expert time.
method SLEEPER uses convolutional neural networks and expert rules to generate interpretable models.
result SLEEPER achieves comparable accuracy to human experts and deep neural networks.

A probabilistic method for deep embedding that improves classification accuracy and interpretability.

problem Improving classification accuracy and interpretability in deep learning.
method A probabilistic approach that treats embeddings as random variables, using a product distribution over labeled instances and marginalizing prototype proximity.
result Superior large- and open-set classification accuracy compared to state-of-the-art methods.

FiberNet integrates geometry into machine learning for clearer classification.

problem Lack of interpretability in traditional deep learning.
method Reformulates classification as geometric optimization on fiber bundles, introducing learnable Riemannian metrics and variational prototype optimization.
result Clear geometric interpretability and efficiency in classification.

Analysis of SGD for Gaussian mixture classification using dynamical mean-field theory.

problem Learning dynamics of SGD for a neural network classifying Gaussian mixture.
method Applying dynamical mean-field theory to track SGD dynamics in high dimensions.
result Reveals how SGD navigates the non-convex loss landscape.

Prototypal analysis is introduced to overcome two shortcomings of archetypal analysis: its sensitivity to outliers and its non-locality, which reduces its applicability as a learning tool. Same as archetypal analysis, prototypal analysis finds prototypes through convex combination of the data points and approximates th…

2017-01-31abs ↗pdf ↗

Bayesian meta-learning on relation graphs improves few-shot relation extraction.

problem Predicting relations in sentences with limited labeled examples.
method Bayesian meta-learning on a global relation graph, using graph neural networks and Langevin dynamics.
result Framework effectively learns and generalizes to new relations.

Method generates prototypes from small datasets for efficient learning.

problem Efficiently learning from small datasets with soft labels.
method Modular method for generating soft-label prototypical lines and Hierarchical Soft-Label Prototype k-Nearest Neighbor algorithm.
result High classification accuracy with significantly fewer prototypes than classes.