AutoML, representation learning and other ideas relevant to crypto-asset predictions.
Earlier this week, I presented a webinar where we discussed some of the new ideas about our work building predictive models and quant strategies for crypto-asset predictions. The session was a big heavy of the machine learning content but tailored to the crypto space.
Here are some of the key ideas we discussed.
· Financial time-series forecasting is incredibly challenging and crypto brings its own set of difficulties.
· Deep learning represents the best opportunity for building robust predictive models for crypto assets.
· Most time-series forecasting frameworks remained very limited to tackle crypto datasets.
· Representation learning is an effective way to streamline feature generation.
· AutoML techniques offer an interesting promise when comes to predictive models.
The slide deck and video can be found below: