Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1391-0
K. Ozaki, S. Nishida, Shinji Nishimoto, H. Asoh, I. Kobayashi
It is known that primary visual cortex uses a sparse code to efficiently represent natural scenes. Based on this fact, we built up a hypothesis that the same phenomenon happens at the higher cognitive function. Here we focus on semantic representation reflecting the meaning of words in the cerebral cortex. We applied sparse coding to the matrix consisting of paired data for both brain activity evoked by visual stimuli observed while a subject is watching a video, and distributed semantic representation made from the description of the video by means of a word2vec language model. Using this method, we obtained a dictionary matrix whose bases represent the corresponding relation between brain activity and the semantic representation. We then analyzed the characteristics of each base in the dictionary matrix. As a result, we confirmed that independent perceptual units were extracted with words representing their functional meaning.
{"title":"Analysis of Correspondence Relationship between Brain Activity and Semantic Representation","authors":"K. Ozaki, S. Nishida, Shinji Nishimoto, H. Asoh, I. Kobayashi","doi":"10.32470/ccn.2019.1391-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1391-0","url":null,"abstract":"It is known that primary visual cortex uses a sparse code to efficiently represent natural scenes. Based on this fact, we built up a hypothesis that the same phenomenon happens at the higher cognitive function. Here we focus on semantic representation reflecting the meaning of words in the cerebral cortex. We applied sparse coding to the matrix consisting of paired data for both brain activity evoked by visual stimuli observed while a subject is watching a video, and distributed semantic representation made from the description of the video by means of a word2vec language model. Using this method, we obtained a dictionary matrix whose bases represent the corresponding relation between brain activity and the semantic representation. We then analyzed the characteristics of each base in the dictionary matrix. As a result, we confirmed that independent perceptual units were extracted with words representing their functional meaning.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121024926","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1283-0
Michael Pereira, C. Machens, R. Costa, T. Akam
A longstanding question in neuroscience is how animals and humans select actions in complex decision trees. Planning, the evaluation of action sequences by anticipating their outcomes, is thought to coexist in the brain with simpler decision-making strategies, such as habit learning and heuristics. Though planning is often required for optimal choice, for many problems simpler strategies yield similar decisions, making them difficult to disambiguate. The scarcity of behavioral tasks that can dissociate planning from other decision mechanisms while generating rich decision data has hindered our understanding of the neural basis of planning. We developed a novel navigation task in which mice navigate to cued goal locations in a complex maze. A targeted search through the large space of possible maze layouts in that environment maximizes the number of decisions that are informative about the use of planning. Over the course of training mice learn shorter paths to goals, and the individual decisions composing these paths are better accounted for by planning than vector navigation. With hundreds of informative decisions per behavioral session, this paradigm opens the door to the study of the neural basis of route planning.
{"title":"Disambiguating planning from heuristics in rodent spatial navigation","authors":"Michael Pereira, C. Machens, R. Costa, T. Akam","doi":"10.32470/ccn.2019.1283-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1283-0","url":null,"abstract":"A longstanding question in neuroscience is how animals and humans select actions in complex decision trees. Planning, the evaluation of action sequences by anticipating their outcomes, is thought to coexist in the brain with simpler decision-making strategies, such as habit learning and heuristics. Though planning is often required for optimal choice, for many problems simpler strategies yield similar decisions, making them difficult to disambiguate. The scarcity of behavioral tasks that can dissociate planning from other decision mechanisms while generating rich decision data has hindered our understanding of the neural basis of planning. We developed a novel navigation task in which mice navigate to cued goal locations in a complex maze. A targeted search through the large space of possible maze layouts in that environment maximizes the number of decisions that are informative about the use of planning. Over the course of training mice learn shorter paths to goals, and the individual decisions composing these paths are better accounted for by planning than vector navigation. With hundreds of informative decisions per behavioral session, this paradigm opens the door to the study of the neural basis of route planning.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125047362","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1411-0
S. B. Yoo, Ł. Kuśmierz, B. Hayden, Taro Toyoizumi
In ecological setup, a wide variety of organisms search over space to obtain reward using information obtained via multiple senses. In the simplest scenario of scalar search, a single quantity, e.g. concentration of a chemoattractant, is measured at different locations. Though gradient is considered a crucial component of scalar search, whether organisms rely solely on the gradient is unknown. We hypothesized that scalar search benefits from information other than gradient, including curvature (second-order derivatives) and long-term memory information integration. To test our hypothesis, we devised an information foraging task. In our task, human subjects control a circular avatar to find a peak of the contour by making brief fixations. They were rewarded when they approached the peak within the predefined maximum number of fixations. In our preliminary data, observed search trajectories deviated from what is expected from the gradient-based search, suggesting that the subjects utilized information beyond the gradient. We also manipulated the perception and action components of the task to examine the sensitivity of the adopted strategies to variations of the task design.
{"title":"Can brains perform second-order optimization?","authors":"S. B. Yoo, Ł. Kuśmierz, B. Hayden, Taro Toyoizumi","doi":"10.32470/ccn.2019.1411-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1411-0","url":null,"abstract":"In ecological setup, a wide variety of organisms search over space to obtain reward using information obtained via multiple senses. In the simplest scenario of scalar search, a single quantity, e.g. concentration of a chemoattractant, is measured at different locations. Though gradient is considered a crucial component of scalar search, whether organisms rely solely on the gradient is unknown. We hypothesized that scalar search benefits from information other than gradient, including curvature (second-order derivatives) and long-term memory information integration. To test our hypothesis, we devised an information foraging task. In our task, human subjects control a circular avatar to find a peak of the contour by making brief fixations. They were rewarded when they approached the peak within the predefined maximum number of fixations. In our preliminary data, observed search trajectories deviated from what is expected from the gradient-based search, suggesting that the subjects utilized information beyond the gradient. We also manipulated the perception and action components of the task to examine the sensitivity of the adopted strategies to variations of the task design.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125834667","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1261-0
Kevin Luxem, Falko Fuhrmann, S. Remy, Pavol Bauer
Recording of neuronal population activity in behaving animals is becoming increasingly popular. Computational markerless annotation tools allow for tracking of animal body-parts throughout the experiment. However, the question remains of how to cross-correlate the extracted behavioral data with the simultaneously acquired neuronal population activity, when both datasets are of high dimensionality. Here we propose a combined analysis, where the behavioral data is clustered into discrete states using a deep learning model and the occurrence of each state is correlated to clusters of neuronal activity. We then model the relationship between behavioral states as a network, where related states are hierarchically grouped while the similarity between their neuronal correlates is maximized. This type of analysis allows for hierarchical exploration of the bidirectional relationship between behavior and its neuronal correlates at different temporal scales.
{"title":"Hierarchical network analysis of behavior and neuronal population activity","authors":"Kevin Luxem, Falko Fuhrmann, S. Remy, Pavol Bauer","doi":"10.32470/ccn.2019.1261-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1261-0","url":null,"abstract":"Recording of neuronal population activity in behaving animals is becoming increasingly popular. Computational markerless annotation tools allow for tracking of animal body-parts throughout the experiment. However, the question remains of how to cross-correlate the extracted behavioral data with the simultaneously acquired neuronal population activity, when both datasets are of high dimensionality. Here we propose a combined analysis, where the behavioral data is clustered into discrete states using a deep learning model and the occurrence of each state is correlated to clusters of neuronal activity. We then model the relationship between behavioral states as a network, where related states are hierarchically grouped while the similarity between their neuronal correlates is maximized. This type of analysis allows for hierarchical exploration of the bidirectional relationship between behavior and its neuronal correlates at different temporal scales.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"54 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125297232","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1112-0
L. Doumas, Guillermo Puebla, J. Hummel, Andrea E. Martin
Humans readily generalize, applying prior knowledge to novel situations and stimuli. Advances in machine learning have begun to approximate and even surpass human performance, but these systems struggle to generalize what they have learned to untrained situations. We present a model based on well-established neurocomputational principles that demonstrates human-level generalization. This model is trained to play one video game (Breakout) and performs one-shot generalization to a new game (Pong) with different characteristics. The model generalizes because it learns structured representations that are functionally symbolic (viz., a role-filler binding calculus) from unstructured training data. It does so without feedback, and without requiring that structured representations are specified a priori. Specifically, the model uses neural co-activation to discover which characteristics of the input are invariant and to learn relational predicates, and oscillatory regularities in network firing to bind predicates to arguments. To our knowledge, this is the first demonstration of human-like generalization in a machine system that does not assume structured representations to begin with.
{"title":"Predicate learning via neural oscillations supports one-shot generalization between video games","authors":"L. Doumas, Guillermo Puebla, J. Hummel, Andrea E. Martin","doi":"10.32470/ccn.2019.1112-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1112-0","url":null,"abstract":"Humans readily generalize, applying prior knowledge to novel situations and stimuli. Advances in machine learning have begun to approximate and even surpass human performance, but these systems struggle to generalize what they have learned to untrained situations. We present a model based on well-established neurocomputational principles that demonstrates human-level generalization. This model is trained to play one video game (Breakout) and performs one-shot generalization to a new game (Pong) with different characteristics. The model generalizes because it learns structured representations that are functionally symbolic (viz., a role-filler binding calculus) from unstructured training data. It does so without feedback, and without requiring that structured representations are specified a priori. Specifically, the model uses neural co-activation to discover which characteristics of the input are invariant and to learn relational predicates, and oscillatory regularities in network firing to bind predicates to arguments. To our knowledge, this is the first demonstration of human-like generalization in a machine system that does not assume structured representations to begin with.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125357374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1192-0
Nathaniel J. Zuk, G. D. Liberto, E. Lalor
The electroencephalographic (EEG) response to a sound of interest is often quantified by averaging time-locked signals over many repetitions in order to get an eventrelated potential (ERP). While this technique can identify an average response, it does not easily allow one to validate the robustness of that response nor variation of the response over repetitions of the sound. Here, we extend the ERP technique as a linear-nonlinear Bernoulli (LNB) model, inspired by neural models, in order to develop a framework for decoding the timing of stimulus events. We use this technique to analyze EEG recordings during presentations of continuous speech and examine neural responses to phonemes, which have been shown to have characteristic EEG responses. Pattern analysis of the confusion between phonemes separates phonemes into vowel and constants, indicating separate ERPs that can robustly predict these phoneme classes. We also find that vowels are decoded more accurately than consonants, and the time course of vowel predictability tracks the rhythm of vowels, while consonant predictability does not track the rhythm of consonants. Overall, we demonstrate a specific instance in which a linear-nonlinear Bernoulli modeling framework can be used to compare ERPs and quantify the ability to decode stimulus events from EEG.
{"title":"Linear-nonlinear Bernoulli modeling for quantifying temporal coding of phonemes in brain responses to continuous speech","authors":"Nathaniel J. Zuk, G. D. Liberto, E. Lalor","doi":"10.32470/ccn.2019.1192-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1192-0","url":null,"abstract":"The electroencephalographic (EEG) response to a sound of interest is often quantified by averaging time-locked signals over many repetitions in order to get an eventrelated potential (ERP). While this technique can identify an average response, it does not easily allow one to validate the robustness of that response nor variation of the response over repetitions of the sound. Here, we extend the ERP technique as a linear-nonlinear Bernoulli (LNB) model, inspired by neural models, in order to develop a framework for decoding the timing of stimulus events. We use this technique to analyze EEG recordings during presentations of continuous speech and examine neural responses to phonemes, which have been shown to have characteristic EEG responses. Pattern analysis of the confusion between phonemes separates phonemes into vowel and constants, indicating separate ERPs that can robustly predict these phoneme classes. We also find that vowels are decoded more accurately than consonants, and the time course of vowel predictability tracks the rhythm of vowels, while consonant predictability does not track the rhythm of consonants. Overall, we demonstrate a specific instance in which a linear-nonlinear Bernoulli modeling framework can be used to compare ERPs and quantify the ability to decode stimulus events from EEG.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"105 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115245754","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1174-0
Avital Hahamy, Timothy Edward John Behrens
Understanding how the brain encodes information is one of the core questions in cognitive neuroscience. This question has been tackled by measuring finegrained fMRI activity patterns across voxels, termed brain representations. These measured representations likely capture gross variations in activity across functional sub-regions, which are reflected in patterns of low spatial frequency. However, it is unclear whether patterns that are not driven by functional/anatomical structure (and are therefore expected to contain higher spatial frequencies) also contribute to these representations. Such rugged patterns have the potential to reflect more intricate stimulus-related information. Here we present a novel method for separating the highfrom the low-frequency patterns, and evaluating whether these patterns contain reliable information. By relying on cross-subject temporal synchronization of brain activity and within-subject consistency of activity patterns, our method provides evidence that, at least in sensory brain regions, highfrequency patterns hold reliable information. Using the same method we also demonstrate that many of these activity patterns are unique to each individual. These results demonstrate the potential of our novel method to shed new light on the types of information conveyed by brain representation.
{"title":"Measuring the Spatial Scale of Brain Representations","authors":"Avital Hahamy, Timothy Edward John Behrens","doi":"10.32470/ccn.2019.1174-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1174-0","url":null,"abstract":"Understanding how the brain encodes information is one of the core questions in cognitive neuroscience. This question has been tackled by measuring finegrained fMRI activity patterns across voxels, termed brain representations. These measured representations likely capture gross variations in activity across functional sub-regions, which are reflected in patterns of low spatial frequency. However, it is unclear whether patterns that are not driven by functional/anatomical structure (and are therefore expected to contain higher spatial frequencies) also contribute to these representations. Such rugged patterns have the potential to reflect more intricate stimulus-related information. Here we present a novel method for separating the highfrom the low-frequency patterns, and evaluating whether these patterns contain reliable information. By relying on cross-subject temporal synchronization of brain activity and within-subject consistency of activity patterns, our method provides evidence that, at least in sensory brain regions, highfrequency patterns hold reliable information. Using the same method we also demonstrate that many of these activity patterns are unique to each individual. These results demonstrate the potential of our novel method to shed new light on the types of information conveyed by brain representation.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122665010","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1148-0
S. Glasauer, Zhuanghua Shi
Perceptual biases found experimentally are often taken to indicate that we should be cautious about the veridicality of our perception in everyday life. Here we show, to the contrary, that such biases may be a consequence of the experimental protocol that cannot be generalized to other situations. We show that the central tendency, an overestimation of small magnitudes and underestimation of large ones, strongly depends on stimulus order. If the same set of stimuli is, rather than being presented in the usual randomized order, is applied in an order that displays only small changes from one trial to the next, the central tendency decreases significantly. This decrease is predicted by a probabilistic model that assumes iterative trial-wise updating of a prior of the stimulus distribution. We conclude that the commonly used randomization of stimuli introduces systematic perceptual biases that may not relevant in everyday life.
{"title":"Central Tendency as Consequence of Experimental Protocol","authors":"S. Glasauer, Zhuanghua Shi","doi":"10.32470/ccn.2019.1148-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1148-0","url":null,"abstract":"Perceptual biases found experimentally are often taken to indicate that we should be cautious about the veridicality of our perception in everyday life. Here we show, to the contrary, that such biases may be a consequence of the experimental protocol that cannot be generalized to other situations. We show that the central tendency, an overestimation of small magnitudes and underestimation of large ones, strongly depends on stimulus order. If the same set of stimuli is, rather than being presented in the usual randomized order, is applied in an order that displays only small changes from one trial to the next, the central tendency decreases significantly. This decrease is predicted by a probabilistic model that assumes iterative trial-wise updating of a prior of the stimulus distribution. We conclude that the commonly used randomization of stimuli introduces systematic perceptual biases that may not relevant in everyday life.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"138 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114423365","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1325-0
Arielle S. Keller, Shi Qiu, Jason Li, L. Williams
Attention impairments are a debilitating symptom of Major Depressive Disorder, yet the neurobiological mechanisms underlying this cognitive dysfunction are poorly understood. Moreover, we currently have no method for predicting how individuals’ attention function may change with antidepressant treatment. Our goal was twofold: First, we modeled the effects of both stress and neural factors implicated in attention impairments and their interactions. To do so, we leveraged a large sample of depressed individuals from the international Study to Predict Optimized Treatment for Depression (iSPOT-D) assessed for attention impairments using a behavioral test, for stress using history of early life stress exposure, and for neural function using electroencephalography (EEG). Second, we developed models for predicting whether attention function changes over time as a function of an eight-week course of antidepressant treatment. Our models demonstrate that 1) early life stress interacts with oscillatory EEG signals to produce attention impairment, and 2) gradient boosted trees can be leveraged to predict changes in attention behavior with treatment. Our models provide novel insight into potential biomarkers of attention impairments in depressed individuals as well as how these impairments may change over time.
{"title":"Modeling attention impairments in major depression","authors":"Arielle S. Keller, Shi Qiu, Jason Li, L. Williams","doi":"10.32470/ccn.2019.1325-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1325-0","url":null,"abstract":"Attention impairments are a debilitating symptom of Major Depressive Disorder, yet the neurobiological mechanisms underlying this cognitive dysfunction are poorly understood. Moreover, we currently have no method for predicting how individuals’ attention function may change with antidepressant treatment. Our goal was twofold: First, we modeled the effects of both stress and neural factors implicated in attention impairments and their interactions. To do so, we leveraged a large sample of depressed individuals from the international Study to Predict Optimized Treatment for Depression (iSPOT-D) assessed for attention impairments using a behavioral test, for stress using history of early life stress exposure, and for neural function using electroencephalography (EEG). Second, we developed models for predicting whether attention function changes over time as a function of an eight-week course of antidepressant treatment. Our models demonstrate that 1) early life stress interacts with oscillatory EEG signals to produce attention impairment, and 2) gradient boosted trees can be leveraged to predict changes in attention behavior with treatment. Our models provide novel insight into potential biomarkers of attention impairments in depressed individuals as well as how these impairments may change over time.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129627874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 1900-01-01DOI: 10.32470/ccn.2019.1414-0
Nicholas M. Blauch, M. Behrmann, D. Plaut
Human expertise for recognizing unfamiliar faces has recently been called into question, highlighting a deficit when compared to familiar face recognition. We present simulations of a fixed-architecture deep convolutional neural network (DCNN) with different training regimens, highlighting the extent to which learning to recognize many "familiar" faces allows for robust, but incomplete, generalization to new "unfamiliar" faces as compared to performance after familiarization. With some training, verification performance for previously unfamiliar faces improves modestly, but the performance difference between unfamiliar and familiar faces is much smaller than the performance boost from pre-training on faces as compared to objects in the ImageNet 1000-way image classification database. We also assess the generalization performance of our networks to other fine-grained visual tasks such as bird species and car model verification. We find that expert face recognition does not improve generalization to birds or cars compared to a network trained on a subset of ImageNet with all vehicles and birds removed. We conclude that the specific learned statistics within a domain of visual expertise determine its generalization to other domains, in contrast with domain-general accounts which highlight level of processing over domain-specific statistics.
{"title":"Visual Expertise and the Familiar Face Advantage","authors":"Nicholas M. Blauch, M. Behrmann, D. Plaut","doi":"10.32470/ccn.2019.1414-0","DOIUrl":"https://doi.org/10.32470/ccn.2019.1414-0","url":null,"abstract":"Human expertise for recognizing unfamiliar faces has recently been called into question, highlighting a deficit when compared to familiar face recognition. We present simulations of a fixed-architecture deep convolutional neural network (DCNN) with different training regimens, highlighting the extent to which learning to recognize many \"familiar\" faces allows for robust, but incomplete, generalization to new \"unfamiliar\" faces as compared to performance after familiarization. With some training, verification performance for previously unfamiliar faces improves modestly, but the performance difference between unfamiliar and familiar faces is much smaller than the performance boost from pre-training on faces as compared to objects in the ImageNet 1000-way image classification database. We also assess the generalization performance of our networks to other fine-grained visual tasks such as bird species and car model verification. We find that expert face recognition does not improve generalization to birds or cars compared to a network trained on a subset of ImageNet with all vehicles and birds removed. We conclude that the specific learned statistics within a domain of visual expertise determine its generalization to other domains, in contrast with domain-general accounts which highlight level of processing over domain-specific statistics.","PeriodicalId":281121,"journal":{"name":"2019 Conference on Cognitive Computational Neuroscience","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129724786","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：