[HKML] Hong Kong Machine Learning Meetup Season 1 Episode 4


  • Wednesday, November 21, 2018 from 6:30 PM to 8:00 PM


  • WeWork Wan Chai 10F Mass Mutual Tower, 33 Lockhart Road Wanchai Hong Kong


Simon did an introductory talk about deep learning techniques applied for natural language processing: CNNs, LSTMs, bi-LSTMs, CRF bi-LSTMs are networks often used to label sentences at the word, or even the character level. His presentation slides. Check out his tutorials!

Tan did a visual introduction to Topological Data Analysis (TDA), the application of discrete topology to study point clouds. These techniques allow for a robust description of the point clouds properties at multiple scales via persistence diagrams. Robustness and persistence of patterns at multiple scales are a desirable properties, especially in the case of noisy and highly stochastic financial time series. Tan uses the persistence diagrams as features to a machine learning classifier (say XGBoost) to predict ETFs returns. The TDA features are amongst the most meaningful ones for prediction according to the model. His slides.

Julien presented one of his quant finance papers: PnL Prediction under Extreme Scenarios. His contribution is essentially to see the PnL prediction problem as a numerical integration via the Simpson’s method. He showed that it predicts correctly the PnL under extreme stressed scenarios. Could machine learning help to predict the PnL variation? It is not clear… (very noisy in high dimension with not enough points due to relatively rare events) The slides.