Downsampling from a Data Science Perspective

Downsampling in data science and data processing is as follows (this excludes the DSP, or digital signal processing, technical definition of downsampling - which is similar in spirit but differently defined).

Downsampling involves reducing the number of data points in a data set to enable comparability (sometimes referred to as "balancing the data").  This helps machine learning models avoid bias towards a dominant class.

Various approaches to downsampling (e.g. random downsampling) are described in this IBM article.

IBM's Guide to Small Language Models

IBM have made a guide to SLMs.

Examples SLMs listed as:

• DistilBERT (DistilBERT is Google's groundbreaking BERT model in "distilled" form (hence the name "Distilled BERT"), retaining 97% of BERT's NLU abilities)
• Gemma
• GPT-4o mini
• Granite
• Llama
• Ministral
• Phi

Biztalk and Azure Integration Services

Biztalk was pitched as an application server and an application integration server.

It is designed to be an integrator of different software systems, and automate business processes (you may have a business process that touches on multiple IT systems - so this solution makes perfect sense - deploy a middleware to manage all the interaction).

Biztalk has/had a bunch of adapters including file adapters, FTP, HTTP, SOAP, SMTP, WCF, MSMQ, MQSeries (now IBM MQ) etc.

It's actually a brilliant product name because it captures the concept of what it does so well.

The replacement for Biztalk is Azure Integration Services.

It is interesting to consider the relationship between "AIS" (as defined above in the Azure context) and "RPA" (or "Robotic Process Automation"). The latter focuses on streamlining human-system interactions, and the former focuses on system-to-system interactions.

Machine Unlearning

As a machine learns, so must it unlearn.  

This ability is needed if an LLM ingests copyrighted content or personal data - it must be able to unlearn information it is not permitted to have. This could also apply to fallacious or untrusted data.

IBM in an article have noted the lack of industry wide tools to evaluated the effectiveness of unlearning.

The IBM piece also highlights research by Microsoft on machine unlearning. This also states the problem of the high cost of retraining models (this costly training process is what has spiked demand for GPUs).

A research paper, which styles itself as a "bridge" paper between unlearning research in classification models to unlearning in generative models focusing on the I2I (image-to-image) generation space.

In the IBM article, the writers go on to describe the SPUNGE framework they have developed for machine unlearning (SPUNGE being short for Split, Unlearn, Merge).

The End of Special Sequences - Removal of Trigraphs in C++17

A trigraph is a three-character sequence interpreted as a single character, as per a programming language's specification. (Digraph is a similar concept for two-character sequences).

Keyboard limitations spurred their use i.e. lack of sufficient buttons to encompass the full character set of a language.

The concept of trigraph was initially invented by the ANSI C committee. IBM have historically opposed the removal of trigraphs from C++.

Purify/Quanity - The Rational Story - But How Rational is Rational?

Rational is a software division of IBM that makes the Purify and Quantify products. Their products are built around the concept of Rational Unified Process (TM). Rational was founded by Paul Levy and Mike Devlin in 1981 and sold to IBM in February 2003, for $2.1 billion. On announcement of the deal, IBM SVP Steve Mills remarked: "Rational is an important element of our e-business on-demand initiative...a perfect complement to our existing four brands - WebSphere, DB2, Lotus and Tivoli."

Products Rational is known for:

• Purify
• Quanitfy
• Rational Rose
• Pure coverage

Read more from their library below.

http://www.ibm.com/developerworks/rational/library/298.html