You have a portfolio of 1000 stocks and are running an optimization to minimize tracking error. If I add an L1 penalty on the weights to your objective function instead of an L2 penalty, what intuitively happens to your positions and why?

你的投资组合有1000只股票，目前在跑一个最小化追踪误差的优化模型。如果我在目标函数里加上对权重的L1惩罚项，而不是L2惩罚项，你的持仓会发生什么直观的变化？为什么？