Computation servers for automatic learning of higher level representations

这笔拨款将用于购买两台16核（处理单元）的计算服务器，以进行“学习高级表示和不变变换”项目所需的密集计算机模拟。该项目旨在开发一种方法，通过这种方法，自动适应系统可以从它接收到的低级（感官）信息中发现和学习更有用的、更高层次的表示。允许这样一个系统对复杂的观测现象形成一个更高层次的、不那么肤浅的理解，有望对未来的观测产生更准确的预测。因此，这种方法的成功将为在工程、金融、商业决策和医学等广泛领域的快速应用打开大门，在这些领域，根据过去观察到的数据进行准确预测是非常有用的。

This grant will allow the purchase of two computation servers with 16 cores (processing units) each, necessary to carry out the intensive computer simulations required for the "Learning higher level representations and invariant transformations" project. This project aims at developing methods by which an automated adaptive system can discover and learn more useful, higher level representations from the low level (sensory) informations it receives. Allowing such a system to form a higher level, less superficial understanding of complex observed phenomena, is expected to yield far more accurate predictions of future observations. A success in this approach will thus open the door to immediate applications in a wide range of fields where accurate predictions from past observed data are immensely useful: engineering,  finance, business decision making and medecine.