About The Company
Established To Innovate
All companies have data – but how to analyse it?
Some companies have large amounts, some have small amounts – but they all have data.
The main way business data is typically analysed, if at all, is via:
i. logistic / linear regression, or
ii. tree based algorithms.
The deep learning community, on the other hand, has been primarily focused on three broad areas of research:
i. vision problems (i.e. object classification / location in pictures), or
ii. language models (i.e. language translation, question & answer), or
iii. games playing models (i.e. chess, Go, arcade games)
Deep learning (i.e. neural network) models have made very large gains in these areas over the last ten years, mainly thanks to computing power advancements (Nvidia graphics cards – GPUs) which have enabled research into larger, more complex, algorithms.
Applying deep learning to business data – AIBMod1
In 2017 I had my first exposure to deep learning models.
When I read machine learning research back then, most papers on non-vision or language problems were based on linear regression or tree models, which struck me as a bit strange.
Why should models capable of analysing image patterns or understanding the semantic subtleties within languages not be useful when looking at business data?
Looking for a new challenge, in 2018 I founded AIBMod and began a full-time exercise in understanding:
i. the theory behind deep learning models and associated techniques, and
ii. python coding and the TensorFlow framework
Neural Transformer – the backbone of AIBMod’s approach
Not long before AIBMod was created a recently published research paper called ‘Attention Is All You Need’2 introduced the neural transformer mechanism.
After reading the paper I realised the neural transformer had numerous properties that could benefit the difficult task of analysing business data.
I decided to make the neural transformer the foundation of AIBMod’s research.
I did not realise it at the time, but I was about to go down a road that would lead me to forensically examine and alter numerous deep learning standard approaches to improve the neural transformer’s performance, at least in the context of business data.
Some companies have large amounts, some have small amounts – but they all have data.
The main way business data is typically analysed, if at all, is via:
i. logistic / linear regression, or
ii. tree based algorithms.
The deep learning community, on the other hand, has been primarily focused on three broad areas of research:
i. vision problems (i.e. object classification / location in pictures), or
ii. language models (i.e. language translation, question & answer), or
iii. games playing models (i.e. chess, Go, arcade games)
Deep learning (i.e. neural network) models have made very large gains in these areas over the last ten years, mainly thanks to computing power advancements (Nvidia graphics cards – GPUs) which have enabled research into larger, more complex, algorithms.
Applying deep learning to business data – AIBMod1
In 2017 I had my first exposure to deep learning models.
When I read machine learning research back then, most papers on non-vision or language problems were based on linear regression or tree models, which struck me as a bit strange.
Why should models capable of analysing image patterns or understanding the semantic subtleties within languages not be useful when looking at business data?
Looking for a new challenge, in 2018 I founded AIBMod and began a full-time exercise in understanding:
i. the theory behind deep learning models and associated techniques, and
ii. python coding and the TensorFlow framework
Neural Transformer – the backbone of AIBMod’s approach
Not long before AIBMod was created a recently published research paper called ‘Attention Is All You Need’2 introduced the neural transformer mechanism.
After reading the paper I realised the neural transformer had numerous properties that could benefit the difficult task of analysing business data.
I decided to make the neural transformer the foundation of AIBMod’s research.
I did not realise it at the time, but I was about to go down a road that would lead me to forensically examine and alter numerous deep learning standard approaches to improve the neural transformer’s performance, at least in the context of business data.
Please continue to browse the site to further understand the research I undertook and the services offered by AIBMod.