HyperShot: Few-Shot Learning by Kernel HyperNetworks

Sendera, Marcin; Przewiȩźlikowski, Marcin; Karanowski, Konrad; Zięba, Maciej; Tabor, Jacek; Spurek, Przemysław

doi:10.1109/wacv56688.2023.00250

Cited by 14 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• • Non-Gaussian Gaussian process (NGGP). 42 Expands on basic Gaussian process techniques for few-shot learning by modeling the posterior distribution with an ODE-based normalizing flow. • MetaNet.…”

Section: • Molbert + Attentive Neural Process (Anp) Combinesmentioning

confidence: 99%

Ligand-Based Compound Activity Prediction via Few-Shot Learning

Eckmann,

Anderson,

et al. 2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

Predicting the activities of new compounds against biophysical or phenotypic assays based on the known activities of one or a few existing compounds is a common goal in early stage drug discovery. This problem can be cast as a "few-shot learning" challenge, and prior studies have developed few-shot learning methods to classify compounds as active versus inactive. However, the ability to go beyond classification and rank compounds by expected affinity is more valuable. We describe Few-Shot Compound Activity Prediction (FS-CAP), a novel neural architecture trained on a large bioactivity data set to predict compound activities against an assay outside the training set, based on only the activities of a few known compounds against the same assay. Our model aggregates encodings generated from the known compounds and their activities to capture assay information and uses a separate encoder for the new compound whose activity is to be predicted. The new method provides encouraging results relative to traditional chemical-similarity-based techniques as well as other state-of-the-art few-shot learning methods in tests on a variety of ligand-based drug discovery settings and data sets. The code for FS-CAP is available at https://github.com/Rose-STL-Lab/FS-CAP.

show abstract

Section: • Molbert + Attentive Neural Process (Anp) Combinesmentioning

confidence: 99%

Ligand-Based Compound Activity Prediction via Few-Shot Learning

Eckmann,

Anderson,

et al. 2024

J. Chem. Inf. Model.

View full text Add to dashboard Cite

show abstract

“…where f −1 β (y) = z. CNFs are rather designed to model complex probability distributions for low-dimensional data, what was confirmed in various applications including point cloud generation [27], future prediction [29] or probabilistic few-shot regression [23]. Compared to models like RealNVP [5] or Glow [11], they can be successfully applied to one-dimensional data and achieve better results for tabular datasets.…”

Section: Decision Tree Ensembles Decision Treementioning

confidence: 99%

“…The low-dimensional representation w of the sparse embedding o is further passed to the conditional CNF module as a conditioning factor. We postulate to use the variant of the conditional flow-based model provided in [27,23], where w is delivered to the function of dynamics, g β (z(t), t, w). The transformation function is given by eq.…”

Section: Treeflowmentioning

confidence: 99%

See 1 more Smart Citation

TreeFlow: Going Beyond Tree-Based Parametric Probabilistic Regression

Wielopolski,

Zięba

2023

Frontiers in Artificial Intelligence and Applications

View full text Add to dashboard Cite

The tree-based ensembles are known for their outstanding performance in classification and regression problems characterized by feature vectors represented by mixed-type variables from various ranges and domains. However, considering regression problems, they are primarily designed to provide deterministic responses or model the uncertainty of the output with Gaussian or parametric distribution. In this work, we introduce TreeFlow, the tree-based approach that combines the benefits of using tree ensembles with the capabilities of modeling flexible probability distributions using normalizing flows. The main idea of the solution is to use a tree-based model as a feature extractor and combine it with a conditional variant of normalizing flow. Consequently, our approach is capable of modeling complex distributions for the regression outputs. We evaluate the proposed method on challenging regression benchmarks with varying volume, feature characteristics, and target dimensionality. We obtain the SOTA results for both probabilistic and deterministic metrics on datasets with multi-modal target distributions and competitive results on unimodal ones compared to tree-based regression baselines.

show abstract

“…Kernel hypernetworks combines the hypernetworks paradigm with kernel methods to realize a new strategy that mimics the human way of learning. In 2023, Marcin Sendera et al [67] first proposed this method. This method directly relies on the kernel-based representations of the support examples and a hypernetwork paradigm to create the query set's classification module.…”

Section: Kernel Hypernetworkmentioning

confidence: 99%

Kernel-based Algorithms for Image Classification: A Review

Liang,

Du,

Duan

et al. 2023

Preprint

View full text Add to dashboard Cite

The field of kernel learning methods has made significant strides since their inception in the 1990s. These methods involve mapping data from the original input space to a higher-dimensional feature space using an implicitly determined nonlinear mapping through a kernel function. In the feature space, a linear model is constructed, enabling the fusion of statistical information and the geometric structure of data samples. Kernel methods have demonstrated their effectiveness in various computer vision applications, including remote sensing image analysis, face recognition, and online scene classification. However, traditional kernel methods face challenges in image classification, such as computational complexity and the design of appropriate kernel functions for complex image features. To address these limitations, we explore two approaches: online kernel learning and the integration of deep learning architectures with kernel methods. We present advancements and applications in image classification tasks, aiming to provide a comprehensive overview of kernel learning methods and identify future research directions in this domain.

show abstract

HyperShot: Few-Shot Learning by Kernel HyperNetworks

Cited by 14 publications

References 13 publications

Ligand-Based Compound Activity Prediction via Few-Shot Learning

Ligand-Based Compound Activity Prediction via Few-Shot Learning

TreeFlow: Going Beyond Tree-Based Parametric Probabilistic Regression

Kernel-based Algorithms for Image Classification: A Review

Contact Info

Product

Resources

About