The following are the various generations of AI, listed chronologically from the oldest to newest, along with information on new features and improvements.
What is a generations of AI ?
Each generation of AI identifies a specific machine learning architecture and fixed hyperparameters. Each generation is used to generate many machine learning pipelines on many financial assets.
Generations of AI are independent of the programming languages, platforms, software, databases and libraries used to create the engine behind the machine learning pipeline. If you’re interested in the tech stack and AI engine you can find more here.
Code names are based on the Books of the Bible.
Current Generations of AI
The following generations of AI are already available on this website. New financial assets are supported regularly on each generation.
Genesis Risk Management
- Manage risk on a single financial asset
- No support for cryptocurrencies
- Active 4 years after financial data is available
- Optimized for long-term (7 years and more)
- Transaction rate between 2 to 20 per year
- Compound Annual Growth Rate is used as a key feature in the objective functions of simulation optimization, clustering and stacking
- Sortino Ratio is used as a key feature in the objective function of hyperparameter optimization
- Most predictors are neural networks performing unsupervised learning and clustering
- Ensembles of predictors are performing reinforcement learning and stacking
- Synthesized risk evaluation seeks to optimize the risk/reward ratio over the next month based on data from the past 4 years
Next Generations of AI
The next generations of AI will be available on this website as soon as one consistent set of features is integrated and is providing improved results. Planned features are subject to change.
Genesis Portfolio Management
Exodus Risk Management
- Manage risk on a single financial asset
- NEW – Support for cryptocurrencies
- NEW – Active 6 months after financial data is available
- NEW – Optimized for short-term (18 months and more)
- NEW – Transaction rate may increase
*NEW features compared to Genesis Risk Management
- Simulation optimization to integrate short-term, mid-term and long-term financial data to improve diversity and adaptive potential for a better short-term risk response
- Addition of predictors trained on short-term features for a better short-term risk response
- Resizing of the ensemble of predictors with more predictors from a longer time frame to leverage more diverse past experience and better adapt the risk response to the current short-term features
Leviticus Risk Management
- Manage risk on a single financial asset
- Support for cryptocurrencies
- Active 6 months after financial data is available
- Optimized for short-term (18 months and more)
- NEW – Leverage past experience from other financial assets
- NEW – Improved accuracy of predictions
- NEW – Improved adaptation to market cycles changes
- NEW – Improved handling of opportunities to buy low
*NEW features compared to Exodus Risk Management
- Objective functions to provide more importance to features such as Maximum Drawdown, Downside Risk, Alpha and Sortino Ratio to improve risk reduction
- Addition of predictors such as Regression Neural Networks and Gradient Boosted Trees to improve accuracy of predictions
- Training old predictors on new data to improve behavioral consistency
- Resizing of the ensemble of predictors with more predictors from other financial assets to leverage various experience from different assets
- Use of multiple ensembles with distinct objective functions and evolution chains adapted to specific market cycles
- Synthesized risk evaluation to integrate opportunity cost in risk/reward ratio to catch more opportunities during high risk periods such as flash crash