A Deeper Look at Bongard Problems

Yun, Xinyu; Bohn, Tanner A.; Ling, Charles X.

doi:10.1007/978-3-030-47358-7_54

Cited by 6 publications

(17 citation statements)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Unfortunately, the dataset does not contain relational tasks, so we will not cover them in this article. Yun et al. (2020) worked on actual Bongard problems using deep learning as part of their system but only tested few-shot classification and also not the original task of generating descriptions.…”

Section: Current Research On Deep Learning For Visual Relational Conceptsmentioning

confidence: 99%

Evaluating the progress of deep learning for visual relational concepts

et al. 2021

View full text Add to dashboard Cite

Convolutional neural networks have become the state-of-the-art method for image classification in the last 10 years. Despite the fact that they achieve superhuman classification accuracy on many popular datasets, they often perform much worse on more abstract image classification tasks. We will show that these difficult tasks are linked to relational concepts from cognitive psychology and that despite progress over the last few years, such relational reasoning tasks still remain difficult for current neural network architectures. We will review deep learning research that is linked to relational concept learning, even if it was not originally presented from this angle. Reviewing the current literature, we will argue that some form of attention will be an important component of future systems to solve relational tasks. In addition, we will point out the shortcomings of currently used datasets, and we will recommend steps to make future datasets more relevant for testing systems on relational reasoning.

show abstract

Section: Current Research On Deep Learning For Visual Relational Conceptsmentioning

confidence: 99%

Evaluating the progress of deep learning for visual relational concepts

et al. 2021

View full text Add to dashboard Cite

show abstract

“…Impressively, Phaeaco’s architecture incorporates both low-level perceptual processes working at the pixel- and high-level symbolic analogical reasoning processes. Since then, additional systems have been proposed to solve BPs, including both deep learning neural network (Yun et al, 2020) and more symbolic (Depeweg et al, 2018) approaches…”

Section: Bpsmentioning

confidence: 99%

“…While this framework improves on the previous Phaeaco approach in terms of solved BPs, its lack of dynamics and the absence of any processes that use physical simulation to predict scene changes prevent it from successfully handling most PBPs. The same applies to a more recent approach (Yun et al, 2020), which employs a pretrained convolutional neural network (CNN) for feature generation and substitutes the symbolic grammar by either a one-level classification tree or a regression layer. In addition, the use of features generated by deep artificial neural networks also prevents the autonomous construction of human-readable rules.…”

Section: Bpsmentioning

confidence: 99%

Perception and simulation during concept learning.

Weitnauer

Goldstone

Ritter

2023

Psychological Review

View full text Add to dashboard Cite

A key component of humans' striking creativity in solving problems is our ability to construct novel descriptions to help us characterize novel concepts. Bongard problems (BPs), which challenge the problem solver to come up with a rule for distinguishing visual scenes that fall into two categories, provide an elegant test of this ability. BPs are challenging for both human and machine category learners because only a handful of example scenes are presented for each category, and they often require the open-ended creation of new descriptions. A new type of BP called physical Bongard problems (PBPs) is introduced, which requires solvers to perceive and predict the physical spatial dynamics implicit in the depicted scenes. The perceiving and testing hypotheses on structures (PATHS) computational model, which can solve many PBPs, is presented and compared to human performance on the same problems. PATHS and humans are similarly affected by the ordering of scenes within a PBP. Spatially or temporally juxtaposing similar (relative to dissimilar) scenes promotes category learning when the scenes belong to different categories but hinders learning when the similar scenes belong to the same category. The core theoretical commitments of PATHS, which we believe to also exemplify open-ended human category learning, are (a) the continual perception of new scene descriptions over the course of category learning; (b) the context-dependent nature of that perceptual process, in which the perceived scenes establish the context for the perception of subsequent scenes; (c) hypothesis construction by combining descriptions into explicit rules; and (d) bidirectional interactions between perceiving new aspects of scenes and constructing hypotheses for the rule that distinguishes categories.

show abstract

“…The dataset measures the human-level concept learning and reasoning of AI agents with the help of 12 000 matrices. Contrary to the approaches from [59,60], for each problem, the test set is not extracted from the context matrices, but two additional test images are associated with each problem instance (see Figs. 7c-7f).…”

Section: Bongard Problemsmentioning

confidence: 99%

A Review of Emerging Research Directions in Abstract Visual Reasoning

Małkiński¹,

Mańdziuk²

2022

Preprint

View full text Add to dashboard Cite

Visual Reasoning (AVR) problems are commonly used to approximate human intelligence. They test the ability of applying previously gained knowledge, experience and skills in a completely new setting, which makes them particularly well-suited for this task. Recently, the AVR problems have become popular as a proxy to study machine intelligence, which has led to emergence of new distinct types of problems and multiple benchmark sets. In this work we review this emerging AVR research and propose a taxonomy to categorise the AVR tasks along 5 dimensions: input shapes, hidden rules, target task, cognitive function, and main challenge. The perspective taken in this survey allows to characterise AVR problems with respect to their shared and distinct properties, provides a unified view on the existing approaches for solving AVR tasks, shows how the AVR problems relate to practical applications, and outlines promising directions for future work. One of them refers to the observation that in the machine learning literature different tasks are considered in isolation, which is in the stark contrast with the way the AVR tasks are used to measure human intelligence, where multiple types of problems are combined within a single IQ test.

show abstract

A Deeper Look at Bongard Problems

Cited by 6 publications

References 2 publications

Evaluating the progress of deep learning for visual relational concepts

Evaluating the progress of deep learning for visual relational concepts

Perception and simulation during concept learning.

A Review of Emerging Research Directions in Abstract Visual Reasoning

Contact Info

Product

Resources

About