Along with the accelerating urbanization process in China, the problem of urban infrastructure layout has become increasingly prominent. The high density of buildings and the extremely unreasonable distribution of infrastructure make the development face great resistance. This paper reveals the problems in the layout of garden infrastructure by studying and analyzing the theoretical foundations of federal learning and distributed learning and provides an in-depth analysis and elaboration of the problem. The paper uses the shape index and landscape index of green infrastructure (green space, arable land, and water bodies), the average width of roads, road network density, and weighted buildings to conduct a comparative study through the differences in ventilation speed and temperature at different layout garden scales. According to the problems existing in the garden layout in the experimental results, corresponding improvement measures are targeted, and the infrastructure layout of the garden is combined with ecology to make the layout within the garden more suitable.