What are the applications in quantitative finance?

Our ML professor once said that although machine learning research is so hot, more than 90% of applications in the industry still use linear models, especially in the field of Predicative Learning.
I am not well informed, but I guess this statement holds true for quantitative trading. The application of machine learning in quantitative trading is still dominated by regression and possibly some decision trees, but linear models are the absolute mainstay. Especially for traders who mainly look for alpha/alpha signals in market data, the Model Capacity of linear models is rarely enough for them, and it is more profitable to find a high-quality signal than to change to a more complex model. Not to mention that the signal-to-noise ratio of market data is extremely low, and a slightly more complex model runs the risk of overfitting.
Does that mean that other machine learning methods have no application in trading? Not really. For miners whose data sources are not limited to market data, but digging everything (including but not limited to Twitter, Internet traffic, weather, various news media, etc.), linear models are obviously not enough. For example, for Behavioral Strategy, Event Driven Strategy, and Index Arbitrage, since they are not sure how the data are related to each other, they will put some complex and even new research results in ML. Sometimes traders also need to do their own NLP (natural language processing) and CV/PR (pattern recognition), which is even more important for machine learning.
In general, high frequency trading (market making) is still king with linear models, statistical arbitrage is a bit richer, and more general (low to medium frequency) algorithmic/quantitative trading uses more diverse machine learning.