Module 1 - Week 3

Preamble

Beyond the academic study and practical learning I am really enjoying getting to know my fellow students.

As I am letting the train take the strain of commuting in and out of central Portsmouth I arrive early at the faculty and it's a great chance to chat with others - something I would have avoided before getting the implant.

Prior to week three kicking off I had a chat with a couple of the female students about their motivation for attending the bootcamp.

One is a project manager with Prince 2 and agile PM qualifications and over 15 years experience.

I was quite frankly staggered when she said that her experience was counting against her now and that she had learned in the last 6 months of searching for a new contract to play it down.

Are organisations becoming so fickle and drinking too much AI Kool-Aid to favour those who have played around with some AI PM tools over those with credible and verifiable experience?

My next conversation was with an ex-BBC investigative journalist who used to work with fellow journalist and presenter Roger Cook.

I remember Roger's confrontational style in the Cook Report and many have copied his approaches since.

My fellow student is very much despairing of today's social media driven 'post-truth' world and fears what will happen when Trump takes power in January.

A third conversation with someone who works for an interesting company developing 'metaverse' experiences for defence organisations brought back memories of working for QinetiQ - good and bad.

But perhaps the best insight from that exchange was the potential to use avatars with AI learning to help treat mental health issues. Apparently, AI has the ability to listen and learn from a patient to a more accurate degree than a therapist in some scenarios and patients feel more comfortable and open conversing with an avatar.

Practical Session 

Having got further immersed in Python based tools during week two, the third week started with a more non-technical approach to interrogating and manipulating datasets. 

Rather than trees being the theme we moved on to birds - or rather one type of antipodean bird I've never heard of before - a Weka.

Here’s a breakdown of the advantages and disadvantages of using Weka for machine learning:

Advantages

User-Friendly Interface

Weka provides an intuitive GUI, making it accessible for beginners in machine learning.

Wide Range of Algorithms

It includes a comprehensive collection of pre-implemented machine learning algorithms for classification, regression, clustering, association, and feature selection.

Data Preprocessing Tools

Weka supports extensive preprocessing techniques, including data filtering, normalization, and feature selection.

Visualization Tools

Offers tools to visualize datasets, classifiers, and clustering results, helping in exploratory data analysis.

Platform Independent

Being Java-based, Weka runs on most platforms, including Windows, macOS, and Linux.

Extensive Documentation

It has detailed documentation and a large community, making it easier to learn and troubleshoot.

Integration with Other Tools

Weka can be integrated with other software and tools like Python (via libraries such as weka.py) and R, enhancing its usability in larger projects.

Open Source

It is free and open-source, allowing users to customize it for specific needs.

Experimenter and Workflow Tools

Supports reproducible experiments and workflow design for running batch experiments and comparative analyses.

Educational Resource

Widely used in academia for teaching machine learning due to its simplicity and comprehensive toolset.

Disadvantages

Limited Scalability

Weka struggles with very large datasets as it loads all data into memory, making it inefficient for big data applications.

Not Suitable for Real-Time Processing

It is primarily designed for batch processing, not for real-time machine learning tasks.

Limited Customizability of Algorithms

Although many algorithms are implemented, customizing or extending them can be challenging compared to libraries like TensorFlow or PyTorch.

Older Algorithms

Focuses more on traditional machine learning algorithms and lacks robust support for modern deep learning frameworks.

Less Flexibility for Advanced Users

Advanced users may find the GUI limiting, as scripting in Weka is less powerful than coding directly in Python, R, or other languages.

Steeper Learning Curve for CLI Use

While the GUI is user-friendly, using Weka through its command-line interface can be challenging for those unfamiliar with it.

Java Dependency

Being Java-based, it may not appeal to users who primarily work in Python or R ecosystems.

Limited Community for Modern Applications

While it has an active community, it is smaller compared to modern frameworks like TensorFlow, PyTorch, or Scikit-learn.