Published in Towards Data Science·Feb 1Member-onlyTuning RL HyperParameters with Hydra's Optuna sweeperConfigure your Stable-Baselines3 tuning pipeline with ease using Hydra and Optuna — Reinforcement Learning (RL) agents are very susceptible to their hyperparameters. The same agent can go from utterly worthless after training to the top of the leaderboard with the correct hyperparameters! Yet finding the correct combination can be very time-consuming or even futile without the right tools. The main goal of…Reinforcement Learning8 min readReinforcement Learning8 min read
Published in Towards Data Science·Aug 16, 2022Member-onlyReinforcement Learning Intro: Markov Decision ProcessA Markov Decision Process is one of the most fundamental knowledge in Reinforcement Learning. It’s used to represent decision making in optimization problems. The version we present here is the Finite MDP, which analyses discrete time, with discrete action problems, with some amount of stochasticity involved. …Reinforcement Learning9 min readReinforcement Learning9 min read
Published in Towards Data Science·Feb 17, 2022Member-onlyTarget Networks: Slow and Steady Wins the RaceReinforcement Learning is the third family of Machine Learning algorithms, after supervised and unsupervised learning. The aim is to find optimal behavior in an environment through interactions and learning from them. In a perfect world, we can efficiently measure the state of the environment, predict the outcome of each of…Machine Learning5 min readMachine Learning5 min read
Dec 2, 2020Member-onlyImplementing Hawkes Processes in Python using tickHawkes processes are a type of stochastic processes. These are used to model stochastic Point Processes. I have presented this algorithm in more theoretical details in this post. There are few resources that go into great depth about this widely used algorithm. There are fewer yet that cover its use…Python2 min readPython2 min read
Nov 19, 2020Member-onlyA brief introduction to Hawkes ProcessesHawkes processes are a type of stochastic processes. These are used to model stochastic — hear random — Point Processes. Hawkes processes a random and finite series of events that are governed by a probabilistic rule. They were originally created to mitigate a disadvantage in Poisson processes, another type of…Financial Modelling3 min readFinancial Modelling3 min read
