Naive Few-Shot Learning: Uncovering the fluid intelligence of machines (2022-05-24T00:00:00.000000Z)

TL;DR

This paper shows that naive (randomly initialized) deep learning models that are trained on a single SCE with a single optimization step can still solve non-trivial versions of the task relatively well, and extends the findings to solve, without any prior training, real-world anomaly detection tasks in the visual and auditory modalities.

Abstract

In this paper, we aimed to help bridge the gap between human fluid intelligence - the ability to solve novel tasks without prior training - and the performance of deep neural networks, which typically require extensive prior training. An essential cognitive component for solving intelligence tests, which in humans are used to measure fluid intelligence, is the ability to identify regularities in sequences. This motivated us to construct a benchmark task, which we term \textit{sequence consistency evaluation} (SCE), whose solution requires the ability to identify regularities in sequences. Given the proven capabilities of deep networks, their ability to solve such tasks after extensive training is expected. Surprisingly, however, we show that naive (randomly initialized) deep learning models that are trained on a \textit{single} SCE with a \textit{single} optimization step can still solve non-trivial versions of the task relatively well. We extend our findings to solve, without any prior training, real-world anomaly detection tasks in the visual and auditory modalities. These results demonstrate the fluid-intelligent computational capabilities of deep networks. We discuss the implications of our work for constructing fluid-intelligent machines.

Authors

T. Barak

1 papers

Y. Loewenstein

1 papers

Naive Few-Shot Learning: Uncovering the fluid intelligence of machines

TL;DR

Abstract

Authors

References34 items

Emergent analogical reasoning in large language models

Few-shot Visual Reasoning with Meta-analogical Contrastive Learning

Meta-Learning Requires Meta-Augmentation

Language Models are Few-Shot Learners

Self-Trained Deep Ordinal Regression for End-to-End Video Anomaly Detection

Solving Raven's Progressive Matrices with Neural Networks

Meta-Learning without Memorization

Data-Efficient Image Recognition with Contrastive Predictive Coding

The Impact of Adverse Weather Conditions on Autonomous Vehicles: How Rain, Snow, Fog, and Hail Affect the Performance of a Self-Driving Car

Unsupervised Detection of Anomalous Sound Based on Deep Learning and the Neyman–Pearson Lemma

Do Deep Generative Models Know What They Don't Know?

Representation Learning with Contrastive Predictive Coding

Measuring abstract reasoning in neural networks

Innate or Acquired? – Disentangling Number Sense and Early Number Competencies

Learning to Compare: Relation Network for Few-Shot Learning

Learning to Generalize: Meta-Learning for Domain Generalization

A simple neural network module for relational reasoning

Commentary: From ‘sense of number’ to ‘sense of magnitude’ – The role of continuous magnitudes in numerical cognition

Densely Connected Convolutional Networks

Deep Residual Learning for Image Recognition

Very Deep Convolutional Networks for Large-Scale Image Recognition

ImageNet classification with deep convolutional neural networks

A measure of intelligence

Scikit-learn: Machine Learning in Python

Abnormal crowd behavior detection using social force model

Long Short-Term Memory

Components of human intelligence

Component Processes in Analogical Reasoning.

Prospective Learning: Back to the Future

Computer models solving intelligence test problems: Progress and implications

From the SelectedWorks of Marcel Adam Just 1990 What one intelligence test measures : A theoretical account of the processing in the Raven Progressive Matrices Test

Complex Information Processing and Intelligence

Manual for Raven's progressive matrices and vocabulary scales

A Gener-alist Agent

Field of Study