We are proud of the passionate team of experts we have at 4most and we would like to share their motivations and thoughts on their specialist areas with you. This is the first in our series of ‘120 Seconds’, a question and answer session with Fabrizio Russo, Head of Data Science.
1. How did you get interested in Machine Learning?
I became interested in Machine Learning (ML) through studying it as one of my electives in my MSc in Applied Statistics. I wanted to study ML in its original form (Statistical Learning) and understand both the principles and the mechanics of it from the ground up, to give me a different view from the data scientists that use it primarily on a theoretical basis. Drawing parallels to well understood techniques such as regression, is key to putting ML in the context of what it is trying to achieve and looking at why it works differently and often better than traditional techniques.
2. What are the top five benefits of ML?
They are all related to the exploratory aim it fulfils. ML is a collection of greedy algorithms that try and optimise an objective function through trial and error; it is a way of writing lengthy code without actually specifying conditions.
The greatest advantage is that it allows for very deep data exploration (hence “deep learning”)
The space these algorithms explore is non-linear, allowing them to establish relationships that traditional regression models would fail to capture
It allows exploration of a wider variety of data sources given its suitability to incorporate unstructured data
It allows for the use of less distinctively predictive variables as it captures nuances instead of average relationships
ML can solve problems when the smallest difference counts the most e.g. anomaly detection.
3. What is the greatest misconception about ML?
That it is a black box. Many ML users gave up on explaining what it is because of its complexity, but this doesn’t mean that it’s not possible; it is just somewhat more difficult. A consequence of this misconception is the reluctance of adopting a ML solution because of the inability to understand and monitor it. Whilst ML interpretability is a relatively unexplored field it has been gaining more attention recently. Some model-agnostic as well as model-specific techniques have already been developed with many others in progress, which is going to be a huge advantage. This is something that 4most is actively working on to solve.
4. What is the hottest ML trend you see happening this year?
The biggest trend I see is ML being implemented in different contexts. I think the evolution in computational power and its speed of change - that makes good infrastructure solutions cheaper - is indicative of the exploration capabilities that companies will have at their disposal. Ultimately, I believe that even the most sceptical will get on board and embrace the analytical revolution.
5. What advice would you give to a CFO/CRO/CEO considering ML?
Data is one of the most valuable assets we hold, and ML allows us to harness its power and use it to our advantage. This not only relates to the more obvious benefits as mentioned earlier but it is fundamental in being able to make quicker and more informed decisions. Being in control of the outcomes of our decisions and acting on them promptly is key; it will produce better outcomes for customers, essentially resulting in faster and more efficient company growth.
For further information, please contact Fabrizio Russo: firstname.lastname@example.org