- The benefits of anticipation 2007-04-22
Johansson, B. and Balkenius, C. (2006). The benefits of anticipation: an experimental study. In Butz, M., Sigaud, O., Pezzulo, G., and Baldassarre, G. (Eds.), Proceedings of third workshop on anticipatory behavior in adaptive learning systems (ABiALS), Rome.
- Empirical Analysis of Generalization and Learning in XCS with Gradient Descent 2007-10-30
Lanzi P. L., Butz M. V., Goldberg D. E. (2007). GECCO 2007: Genetic and Evolutionary Computation Conference. 1814-1821.
- A developmental approach to dynamic scene understanding 2007-04-22
Balkenius, C., and Johansson, B. (2006). A developmental approach to dynamic scene understanding. Proceedings of the Sixth International Conference on Epigenetic Robotics (p. 165). Lund University Cognitive Studies, 128.
- Complex brain function depends on flexibility 2013-05-19
- Over the past few decades, neuroscientists have made much progress in mapping the brain by deciphering the functions of individual neurons that perform very specific tasks, such as recognizing the location or color of an object.
However, there are many neurons, especially in brain regions that perform sophisticated functions such as thinking and planning, that don’t fit into this pattern. Instead of responding exclusively to one stimulus or task, these neurons react in different ways to a wide variety of things. MIT neuroscientist Earl Miller first noticed these unusual activity patterns about 20 years ago, while recording the electrical activity of neurons in animals that were trained to perform complex tasks.
“We started noticing early on that there are a whole bunch of neurons in the prefrontal cortex that can’t be classified in the traditional way of one message per neuron,” recalls Miller, the Picower Professor of Neuroscience at MIT and a member of MIT’s Picower Institute for Learning and Memory.
In a paper appearing in Nature on May 19, Miller and colleagues at Columbia University report that these neurons are essential for complex cognitive tasks, such as learning new behavior. The Columbia team, led by the study’s senior author, Stefano Fusi, developed a computer model showing that without these neurons, the brain can learn only a handful of behavioral tasks.
“You need a significant proportion of these neurons,” says Fusi, an associate professor of neuroscience at Columbia. “That gives the brain a huge computational advantage.”
Lead author of the paper is Mattia Rigotti, a former grad student in Fusi’s lab.
Miller and other neuroscientists who first identified this neuronal activity observed that while the patterns were difficult to predict, they were not random. “In the same context, the neurons always behave the same way. It’s just that they may convey one message in one task, and a totally different message in another task,” Miller says.
For example, a neuron might distinguish between colors during one task, but issue a motor command under different conditions.
Miller and colleagues proposed that this type of neuronal flexibility is key to cognitive flexibility, including the brain’s ability to learn so many new things on the fly. “You have a bunch of neurons that can be recruited for a whole bunch of different things, and what they do just changes depending on the task demands,” he says.
At first, that theory encountered resistance “because it runs against the traditional idea that you can figure out the clockwork of the brain by figuring out the one thing each neuron does,” Miller says.
For the new Nature study, Fusi and colleagues at Columbia created a computer model to determine more precisely what role these flexible neurons play in cognition, using experimental data gathered by Miller and his former grad student, Melissa Warden. That data came from one of the most complex tasks that Miller has ever trained a monkey to perform: The animals looked at a sequence of two pictures and had to remember the pictures and the order in which they appeared.
During this task, the flexible neurons, known as “mixed selectivity neurons,” exhibited a great deal of nonlinear activity — meaning that their responses to a combination of factors cannot be predicted based on their response to each individual factor (such as one image).
Fusi’s computer model revealed that these mixed selectivity neurons are critical to building a brain that can perform many complex tasks. When the computer model includes only neurons that perform one function, the brain can only learn very simple tasks. However, when the flexible neurons are added to the model, “everything becomes so much easier and you can create a neural system that can perform very complex tasks,” Fusi says.
The flexible neurons also greatly expand the brain’s capacity to perform tasks. In the computer model, neural networks without mixed selectivity neurons could learn about 100 tasks before running out of capacity. That capacity greatly expanded to tens of millions of tasks as mixed selectivity neurons were added to the model. When mixed selectivity neurons reached about 30 percent of the total, the network’s capacity became “virtually unlimited,” Miller says — just like a human brain.
Mixed selectivity neurons are especially dominant in the prefrontal cortex, where most thought, learning and planning takes place. This study demonstrates how these mixed selectivity neurons greatly increase the number of tasks that this kind of neural network can perform, says John Duncan, a professor of neuroscience at Cambridge University.
“Especially for higher-order regions, the data that have often been taken as a complicating nuisance may be critical in allowing the system actually to work,” says Duncan, who was not part of the research team.
Miller is now trying to figure out how the brain sorts through all of this activity to create coherent messages. There is some evidence suggesting that these neurons communicate with the correct targets by synchronizing their activity with oscillations of a particular brainwave frequency.
“The idea is that neurons can send different messages to different targets by virtue of which other neurons they are synchronized with,” Miller says. “It provides a way of essentially opening up these special channels of communications so the preferred message gets to the preferred neurons and doesn’t go to neurons that don’t need to hear it.”
The research was funded by the Gatsby Foundation, the Swartz Foundation and the Kavli Foundation.
- Toward an integrated biomimetic model of reaching. 2008-05-15
Caligiore D., Parisi D., Baldassarre G. (2007). Towards an integrated biomimetic model of reaching. Demiris Y., Scassellati B., Mareschal D. (eds.). The 6th IEEE International Conference on Development and Learning (ICDL2007). IEEE Catalog Number: 07EX1740C, ISBN: 1-4244-1116-5, Library of Congress: 2007922394, pp. 241-246. London: Imperial College
- RNN-based Learning of Compact Maps for Efficient Robot Localization 2007-04-17
A. Foerster, A. Graves, J. Schmidhuber: RNN-based Learning of Compact Maps for Efficient Robot Localization. ESANN 2007.
- An unforgettable life 2013-05-14
- In 1953, a young man named Henry Molaison underwent an experimental operation that doctors hoped would control his frequent epileptic seizures. When the surgeon could not locate the origin of Molaison’s seizures, he removed a structure known as the hippocampus from both sides of his brain.
Soon after the surgery, Molaison’s doctors realized that the procedure had had a dramatic and unintended consequence: Molaison could no longer form new memories. This tragic loss for Molaison and his family turned him into one of the most important patients in the history of neuroscience. In fact, his case answered more questions about how memory works than the entire previous century of research, writes Suzanne Corkin, MIT professor of neuroscience emerita, in her new book, “Permanent Present Tense.”
“He was a hero,” says Corkin, who devoted much of her career to studying the patient who became world-famous as “H.M.” “He was someone who had something really awful happen to him, and in spite of this, he made a huge contribution.”
In her book, Corkin hopes to shed light on the life of the person behind the scientific studies. “I don’t want people to think of him as just a lesion and a bunch of test scores. He was a real person,” she says. “He had a sense of self, he had values, and he had a wonderful sense of humor.”
Molaison was also aware that he was helping scientists, and was eager to participate in their studies, Corkin says. In a 1992 radio interview, she asked Molaison what he was going to do the next day. “Whatever is beneficial,” he replied.
Breaking down memory
Corkin and Molaison first met in 1962, while Corkin was a graduate student in Brenda Milner’s lab at McGill University. Milner had done the first cognitive evaluation of Molaison in 1955, two years after his operation. Hans-Lukas Teuber recruited Corkin to MIT in 1964, and she set about establishing the behavioral neuroscience laboratory at the MIT Clinical Research Center (CRC). Molaison visited this facility dozens of times to participate in studies designed to probe the mechanisms of human memory.
Milner’s initial studies with Molaison supported the hypothesis that long- and short-term memory recruit different brain processes and regions. His short-term memory was intact: If a researcher read him a string of digits, he could recite it back immediately. As the delay between hearing the numbers and repeating them back increased, however, his performance declined markedly, helping scientists to determine the limits of short-term memory.
Studies with Molaison also helped scientists clarify the distinction between two types of long-term memory now known as declarative and nondeclarative. Declarative memory, also termed explicit memory, refers to conscious memory for facts and specific experiences. Nondeclarative memory, also called procedural or implicit memory, includes skills such as learning to ride a bicycle or play tennis.
When asked to trace the outline of a star by looking at his hand, the pen and the star only as a reflection in a mirror, a task commonly used to evaluate the ability to learn new motor skills, Molaison did gradually improve. Each time he performed the task, however, he had no recollection of having done it before. This revealed that Molaison’s nondeclarative memory was still intact, even though he lacked the declarative knowledge of having learned something new.
To Corkin’s great surprise, Molaison was also able to draw, in 1966, the floor plan of the house where he was living at the time, where he had moved after his operation. “He had no business knowing anything about the house, but over several years he had built up a representation of the floor plan,” Corkin says. She believes that brain areas such as the cortex and memory structures located near the hippocampus could still perform some memory functions and gradually absorbed knowledge of the house’s layout.
These findings and other studies suggest that there are many aspects of memory distributed throughout the brain, and that Molaison was able to tap into some of those areas that were still intact.
Part of the family
Despite Molaison’s amnesia, he did seem to know something about what had happened to him, Corkin says. “He knew he had epilepsy, he knew he had an operation, and he knew he had a memory impairment,” she says. “He also had a vague sense that the operation had been done only a few times before, and that when it was done with him something went wrong.”
After his operation, Molaison lost access to specific events from his childhood, although he remembered the gist of his early experiences. He was able to recognize family members and celebrities whom he had known or heard of before 1953.
Over the decades that Corkin knew Molaison, he became much more than a research subject, she says. “He was part of our lab family,” she says. “We celebrated his 60th birthday at the CRC. We would send him gifts at Christmastime or other times. If we knew he needed something we’d send it. He loved Westerns, so we sent him a bunch of those, we sent him the equipment to play movies on.”
Molaison eventually started to believe that he knew Corkin, but thought that she had been a high school classmate. He also got used to seeing the nurses and other patients in the nursing home where he spent the last 28 years of his life, though he didn’t remember any of their names.
In 1992, Molaison donated his brain to MIT and Massachusetts General Hospital for scientific study after his death. After he died, on Dec. 2, 2008, at the age of 82, his brain was scanned both at Mass General and at the University of California at San Diego. One year after that, researchers cut his brain into 2,401 slices, each as thin as a human hair, to create a digital map of it. A live stream of the two-day process was viewed more than 3 million times.
“He is so much more famous than all of the scientists who studied him put together,” Corkin says. “He revolutionized the science of memory. We look at memory now in a totally different way than people did before H.M.”
- Learning to anticipate a temporarily hidden moving object 2007-11-13
- How Can a Massively Modular Mind Be Context-Sensitive? A Computational Approach 2007-04-15
Giovanni Pezzulo (2006). Proceedings of 7th International Conference on Cognitive Modeling (ICCM 2006).