an interesting review of work by Desrochers et al.
, which examines whether basic principles of reinforcement learning, coupled with a complex environment and a large memory, might account for more complex behaviors. They show that reinforcement learning can explain not only behavioral choice in a complex environment, but also the evolution toward optimal behavior over a long time. They studied, in the monkey, the sort of eye movements we make several times a second when scanning a complex image (the scan path is dramatically influenced by what we are thinking.) Here is their abstract, followed by Sejnowski's summation.
Habits and rituals are expressed universally across animal species. These behaviors are advantageous in allowing sequential behaviors to be performed without cognitive overload, and appear to rely on neural circuits that are relatively benign but vulnerable to takeover by extreme contexts, neuropsychiatric sequelae, and processes leading to addiction. Reinforcement learning (RL) is thought to underlie the formation of optimal habits. However, this theoretic formulation has principally been tested experimentally in simple stimulus-response tasks with relatively few available responses. We asked whether RL could also account for the emergence of habitual action sequences in realistically complex situations in which no repetitive stimulus-response links were present and in which many response options were present. We exposed naïve macaque monkeys to such experimental conditions by introducing a unique free saccade scan task. Despite the highly uncertain conditions and no instruction, the monkeys developed a succession of stereotypical, self-chosen saccade sequence patterns. Remarkably, these continued to morph for months, long after session-averaged reward and cost (eye movement distance) reached asymptote. Prima facie, these continued behavioral changes appeared to challenge RL. However, trial-by-trial analysis showed that pattern changes on adjacent trials were predicted by lowered cost, and RL simulations that reduced the cost reproduced the monkeys’ behavior. Ultimately, the patterns settled into stereotypical saccade sequences that minimized the cost of obtaining the reward on average. These findings suggest that brain mechanisms underlying the emergence of habits, and perhaps unwanted repetitive behaviors in clinical disorders, could follow RL algorithms capturing extremely local explore/exploit tradeoffs.
Sejnowski's review gives several other examples of reinforcement learning solving difficult problems (such as learning how to play Blackgammon), and concludes:
...the jury is still out on whether reinforcement learning can explain the highest levels of human achievement. Rather than add a radically new piece of machinery to the brain, such as a language module, nature may have tinkered with the existing brain machinery to make it more efficient. Children have a remarkable ability to learn through imitation and shared attention, which might greatly speed up reinforcement learning by focusing learning on important stimuli. We are also exceptional at waiting for rewards farther into the future than other species, in some cases delaying gratification to an imagined afterlife made concrete by words. Supercharged with a larger cerebral cortex, faster learning, and a longer time horizon, is it possible that we solve complex problems in mathematics the same way that monkeys find optimal scan paths?
Post a Comment