Rolls E T, Treves A, Tovee M J
University of Oxford, Department of Experimental Psychology, UK.
Exp Brain Res. 1997 Mar;114(1):149-62. doi: 10.1007/pl00005615.
It has been shown that it is possible to read, from the firing rates of just a small population of neurons, the code that is used in the macaque temporal lobe visual cortex to distinguish between different faces being looked at. To analyse the information provided by populations of single neurons in the primate temporal cortical visual areas, the responses of a population of 14 neurons to 20 visual stimuli were analysed in a macaque performing a visual fixation task. The population of neurons analysed responded primarily to faces, and the stimuli utilised were all human and monkey faces. Each neuron had its own response profile to the different members of the stimulus set. The mean response of each neuron to each stimulus in the set was calculated from a fraction of the ten trials of data available for every stimulus. From the remaining data, it was possible to calculate, for any population response vector, the relative likelihoods that it had been elicited by each of the stimuli in the set. By comparison with the stimuli actually shown, the mean percentage correct identification was computed and also the mean information about the stimuli, in bits, that the population of neurons carried on a single trial. When the decoding algorithm used for this calculation approximated an optimal, Bayesian estimate of the relative likelihoods, the percentage correct increased from 14% correct (chance was 5% correct) with one neuron to 67% with 14 neurons. The information conveyed by the population of neurons increased approximately linearly from 0.33 bits with one neuron to 2.77 bits with 14 neurons. This leads to the important conclusion that the number of stimuli that can be encoded by a population of neurons in this part of the visual system increases approximately exponentially as the number of cells in the sample increases (in that the log of the number of stimuli increases almost linearly). This is in contrast to a local encoding scheme (of "grandmother" cells), in which the number of stimuli encoded increases linearly with the number of cells in the sample. Thus one of the potentially important properties of distributed representations, an exponential increase in the number of stimuli that can be represented, has been demonstrated in the brain with this population of neurons. When the algorithm used for estimating stimulus likelihood was as simple as could be easily implemented by neurons receiving the population's output (based on just the dot product between the population response vector and each mean response vector), it was still found that the 14-neuron population produced 66% correct guesses and conveyed 2.30 bits of information, or 83% of the information that could be extracted with the nearly optimal procedure. It was also shown that, although there was some redundancy in the representation (with each neuron contributing to the information carried by the whole population 60% of the information it carried alone, rather than 100%), this is due to the fact that the number of stimuli in the set was limited (it was 20). The data are consistent with minimal redundancy for sufficiently large and diverse sets of stimuli. The implication for brain connectivity of the distributed encoding scheme, which was demonstrated here in the case of faces, is that a neuron can receive a great deal of information about what is encoded by a large population of neurons if it is able to receive its inputs from a random subset of these neurons, even of limited numbers (e.g. hundreds).
研究表明，仅从一小群神经元的放电率中，就有可能解读猕猴颞叶视觉皮层中用于区分所观看的不同面孔的编码。为了分析灵长类动物颞叶皮质视觉区域单个神经元群体所提供的信息，在一只执行视觉注视任务的猕猴中，分析了14个神经元组成的群体对20种视觉刺激的反应。所分析的神经元群体主要对面孔做出反应，所使用的刺激都是人类和猴子的面孔。每个神经元对刺激集的不同成员都有自己的反应模式。根据每个刺激可用的十次试验数据中的一部分，计算每个神经元对集合中每个刺激的平均反应。从其余数据中，可以为任何群体反应向量计算出它由集合中的每个刺激引发的相对可能性。通过与实际呈现的刺激进行比较，计算出平均正确识别百分比，以及神经元群体在单次试验中携带的关于刺激的平均信息量（以比特为单位）。当用于此计算的解码算法近似于相对可能性的最优贝叶斯估计时，正确百分比从单个神经元时的14%正确（随机概率为5%正确）增加到14个神经元时的67%。神经元群体传达的信息从单个神经元时的0.33比特近似线性地增加到14个神经元时的2.77比特。这得出了一个重要结论，即视觉系统这一部分中的神经元群体能够编码的刺激数量随着样本中细胞数量的增加而近似指数增长（因为刺激数量的对数几乎呈线性增加）。这与“祖母细胞”的局部编码方案形成对比，在该方案中，编码的刺激数量随样本中的细胞数量线性增加。因此，分布式表征的一个潜在重要特性，即能够表征的刺激数量呈指数增加，已在这一神经元群体中得到了证实。当用于估计刺激可能性的算法尽可能简单，易于接收群体输出的神经元实现（仅基于群体反应向量与每个平均反应向量之间的点积）时，仍然发现由14个神经元组成的群体做出了66%的正确猜测，并传达了2.30比特的信息，即几乎最优程序所能提取信息的83%。研究还表明，尽管表征中存在一些冗余（每个神经元对整个群体携带信息的贡献为其单独携带信息的60%，而非100%），但这是由于集合中的刺激数量有限（为20个）。对于足够大且多样的刺激集，数据与最小冗余是一致的。这里在面孔案例中所展示的分布式编码方案对大脑连接性的启示是，如果一个神经元能够从这些神经元的随机子集中接收输入，即使数量有限（例如数百个），它也能够接收到关于大量神经元所编码内容的大量信息。
