Masinõppe kasutamine ajuandmete analüüsiks lühimälu nõudva katse ajal
Date
2014
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Käesolev töö uurib ahvi prefrontaalse ajukoore neuronite elektrilist aktiivsust, kui ahv osaleb lühimälu nõudvas katses. Analüüsi esimeses osas kasutatakse juhendatud masinõpet, et näha, kas aju aktiivsuse põhjal on võimalik ennustada ahvi käitumist. Leitakse, et kuigi enne käitumist ei suudeta seda ennustada, on 69% tõenäosusega võimalik see õigesti välja lugeda käitumisejärgsest neuronite aktiivsusest. Analüüsi teises osas uuritakse, kuidas neuronite aktiivsus sadu katseid sisaldava päeva jooksul muutub. Leitakse, et enamasti püsib aktiivsus sama, kuid osadel neuronitel tõuseb või langeb. Samuti leitakse, et neuroni muutus päeva jooksul ei ole seotud sellega, kuidas see katse jooksul käitub. Need leiud võivad viia parema arusaamiseni prefrontaalse ajukoore ja lühimälu omadustest.
We analyze the electrical activity of neurons in the prefrontal cortex of a monkey while it performs a task requiring short-term memory. In the first part of the analysis we use supervised machine learning to see if we can predict the monkey's behavior from its brain activity. We find that, while unable to predict the behavior before it occurs, we are able to correctly determine it based on post-behavior brain activity 69% of the time. In the second part of the analysis we investigate how the activity of neurons changes during a day of repeating the task hundreds of times. We find that for many neurons it remains the same, but for some it increases or decreases. In addition, we find that how the activity of a neuron changes over the day is not related to how the neuron behaves during the task. These findings can lead to a better understanding of the properties of the prefrontal cortex and short-term memory.
We analyze the electrical activity of neurons in the prefrontal cortex of a monkey while it performs a task requiring short-term memory. In the first part of the analysis we use supervised machine learning to see if we can predict the monkey's behavior from its brain activity. We find that, while unable to predict the behavior before it occurs, we are able to correctly determine it based on post-behavior brain activity 69% of the time. In the second part of the analysis we investigate how the activity of neurons changes during a day of repeating the task hundreds of times. We find that for many neurons it remains the same, but for some it increases or decreases. In addition, we find that how the activity of a neuron changes over the day is not related to how the neuron behaves during the task. These findings can lead to a better understanding of the properties of the prefrontal cortex and short-term memory.