I will share a video featured in the last September 2016 Deep Learning in Finance Summit held in London. This will also serve as a needed interval from the paper review series. I think it to be wise choice. This Blog was not meant to be a daily paper review Blog.
I am also aware of yesterday’s mistakes I’ve made in the paper review post. I was doing the post a little bit too much excited and in a hurry… I will need to be more careful regarding the use of the English Language, will try to read carefully before posting from now on, and also recognize the need to properly read through the papers before posting. That is easier said than done, specially when you feel the rush to post.
So today’s post depart a bit from that stress to present a very nice talk held at the last Deep Learning in Finance Summit. This is form September this year, so it is quite fresh thinking and an update to the state-of-the-art on this subject. The speaker is Peter Sarlin from Hanken School of Economics, a Finnish Economics Higher Education Institution. Peter has extensive experience as a deep learning practitioner in the context of economics and finance having had experience in the subject before in the context of analysis of vulnerable systems in general.
In the video it is presented a link to the paper on systemic risk called RiskRank: Measuring interconnected risk that serves as the underlying background for the points Dr. Peter Sarlin talks about. This is recommended reading. Of note also is the work that Dr. Peter’s company RiskLab together with a financial risk analytics venture firm have done in joint venture with European central bank institutions like the European Central Bank (ECB).
An important remark from this talk is the fact that Dr. Sarlin emphasizes the greater need to a better human-machine interaction understanding. That understanding nowadays isn’t just a question of the practitioner or professional in deep learning implementation/ deployment or the artificial intelligence developer. It is also related with the economic/financial agent in general, that today is more and more a human-machine interaction. But, as it is pointed out, machines will need to improve their understanding of how humans interact with other humans as well.
A final comment on one specific deep learning system in the talk: Almax Analytics – semantic deep learning. This appeared to be a state of the art framework to analyse unstructured news articles form the wide media applied here to the European banking sector. It turns that unstructured data in a supervised dataset able to spot stress (negative impact) levels in that specific article.