Visualising AI Embeddings in APEX

Image
" To deal with hyper-planes in a 14-dimensional space, visualize a 3-D space and say 'fourteen' to yourself very loudly. Everyone does it."     - Geoffrey Hinton , 2018 Turing Award winner. Within the wonderful world of  Generative AI , one concept that is all the rage is RAG, or  Retrieval Augmented Generation , which is an  AI framework that combines the strengths of traditional information retrieval systems (such as databases) with the capabilities of generative large language models (LLM s) . RAG's goal is to improve the accuracy, relevance, and timeliness of information  generation  - such as  documents, text and images -  by optimizing LLM output.  When creating a RAG system, it’s essential to store information in a format that a LLM can retrieve. This is where  data is converted into embeddings through a process that uses pre-trained m
Post a Comment

Fun with Bubbles!

I'm on a train up to Liverpool for the UK Oracle User Group Tech 14 conference, and decided it's about time for a post showing something I've been working on.  I don't post often these days, which I blame on work and family commitments, but is probably just down to pure laziness.  Anyhow, I'm feeling in the mood so I'll just get on with it in case the desire to catch some zzz's becomes overwhelming.  Also it takes my mind of the bad coffee served up on the train.

So, I've been getting into D3 a little, which if you don't already know is a JavaScript library for manipulating data to create interactive graphs and charts.  Some of the stuff people are creating using D3 is amazing, including this dynamic bubble chart example.  I thought I would use this example as the basis for creating a bubble chart in Apex, that dynamically updates with some cool transitions, using data from the database rather than the real time API in the example.

The demo can be found here (login using demo/demo):  https://apex.oracle.com/pls/apex/f?p=9208

Here's a screenshot (not animated sorry, as no time to make one):


Each bubble in the demo is a country, and is coloured according to which region it belongs in.  You can see the regions in the legend to the right.   The size of each bubble is based on a simple value stored for each country, and is not representative of anything in particular (it's server hits on the original demo).  I've stored a number of different values for each country and these are pulled from the DB using an Ajax call (apex.server.process) with a javascript interval which cycles through each set of data.

Interestingly, whilst I was working on this I noticed that Carsten Czarski was also working on some D3 demos (here), and actually these helped me solve an issue I had (I ended up ditching a call to apex_util.json_from_sql as it wasn't correct for this).

The one thing I want add is text labels (country code and value), but I've not quite figured that out yet though will endeavour to do so as it will make it more intuitive.

I'll add another post later with all the code, which is really not that complex.  Promise!

Comments

Unknown said…
I only recently discovered D3 and was amazed by the visualizations. I immediately started experimenting with the library in APEX. Glad to know that others are doing the same. Hoping to share my discoveries later. Good luck on the journey.
8 December 2014 at 12:02
Paul Brookes said…
Thanks Steve. I see the Apex development team use D3, so maybe it could make it into a later release somehow. Also I notice there are a number of Apex/D3 related blog post demos coming out all the time, so definitely worth exploring further.
9 December 2014 at 11:01
Joel R. Kallman said…
Actually, all of the approvals have been completed internally for d3.js to be included in APEX 5.0. Thus, it will "make it into a later release". Let the good times roll.
11 December 2014 at 05:47
Paul Brookes said…
That's great news Joel. With D3 in the mix, Apex 5 will be a serious player in the BI space.
11 December 2014 at 14:07
Post a Comment

Popular posts from this blog

Visualising AI Embeddings in APEX

Image
" To deal with hyper-planes in a 14-dimensional space, visualize a 3-D space and say 'fourteen' to yourself very loudly. Everyone does it."     - Geoffrey Hinton , 2018 Turing Award winner. Within the wonderful world of  Generative AI , one concept that is all the rage is RAG, or  Retrieval Augmented Generation , which is an  AI framework that combines the strengths of traditional information retrieval systems (such as databases) with the capabilities of generative large language models (LLM s) . RAG's goal is to improve the accuracy, relevance, and timeliness of information  generation  - such as  documents, text and images -  by optimizing LLM output.  When creating a RAG system, it’s essential to store information in a format that a LLM can retrieve. This is where  data is converted into embeddings through a process that uses pre-trained m
Read more

Oracle Select AI and APEX

Image
Yes, yes, I know, it's my first blog post in 10 years.  A s Oracle is doing some really interesting stuff with AI I thought I'd break cover and do a post showing a little something I've been working on. It's probably not escaped your attention that Oracle recently announced the release of Oracle Database 23ai (on OCI/Azure now but not yet on-prem), with a host of new features to allow developers to incorporate generative AI (genAI) into their work which will ultimately help end users be more productive. One AI feature that was released in 2023 prior to this announcement is Select AI , or Autonomous Database Select AI to give it the proper title.   This is a feature of the Autonomous Database that enables data to be queried using natural language, generating (ANSI) SQL from genAI large language models (LLMs) such as  OpenAI,  Cohere and Azure OpenAI.   There's a great  demo put together by the Oracle APEX team demonstrating Select AI working with APEX.  In the demo
Read more