//This post was a part of my 2022 Application for The Knowledge Society for its Global Virtual Program
Teach us about a topic you are interested in.
Machine
Learning involves computer algorithms that learn by experience, forming
patterns from a given set of data to determine possible causes that have led to
the current situation.
Neural
networks come under a subset of Machine Learning which is itself a part of the
field Artificial intelligence. Artificial Neural Networks aim to stimulate the
working of the biological mind. The algorithms begin to find patterns from a
given set of data much like the countless deliberations we are required to do before
making a relatively simple decision. (Should I drink Coffee at 9 in the night
or go with a decaf?). We mull over seemingly unrelated clauses which could be
affecting our decision (a friend who is addicted to caffeine) and recount past experiences
(the time when I tossed and turned the whole night) and predict possible
outcomes (I could disrupt my sleep schedule and be unable to make it to school
on time). The truth is that real life is messy and complicated. Neural networks
imitate the thought process that goes into making choices. I often have to analyze data which may seem unrelated
to come to decisions that are not biased (At least not heavily).
Insights
from neuroscience into how neurons function and adapt to change has helped in
forming the core of Machine Learning. Simulated Neural Networks have a number
of layers or nodes present between input and output layers. Research on how
memory is processed, stored and retrieved, and how past experiences influence
learning can help in building smarter machines which can aid us in decision
making, especially in situations that involve many people to be considered,
increasing the influencing factors. Neural Networks could help in making
decisions in the business world. For example, analyzing small businesses which
currently are selling products similar to the established business for cues
that may signal possible disruptive innovation in the future. Neural Networks
are useful as they include input data, weights and bias, similar to how I record the visible
expressions and tone of voice of a friend while in conversation (input), decide
that her shallow breathing and nervous fidgeting signal anxiety and those cues
are more important than the fact she is smiling to reassure me (weights) and derive
the conclusion that her anxiety is due to approaching exams because I regard
her as studious (bias). These processes ultimately influence my actions as I
choose to listen to her thoughts to provide assistance.
Machine
learning could help in building a smart car which involves taking in input (the
traffic on the road, vehicle type, size and acceleration. Inputs could include scanning
the driver’s visible appearance to calculate age and reflex skills to better
predict changes in trajectory) and relying on previously measured patterns to
calculate possible trajectories of vehicles. The software can calculate and store
data if a new pattern emerges. It can rate which calculations are required on a
regular basis, rating the patterns through factors such as time of day, weather
etc. to associate them with situations that require them. Adding in features
such as measuring the friction on the road in cases of rain, hail or snow could
help in regulating the velocity of the car thereby reducing chances of skidding
and a possible accident. The car could record inputs even when a person is
driving it, similar to how a child learns by observing an adult performing the
action. It could analyze other vehicles on the road to measure how they
accelerate, take turns etc.
An enhanced
approach to machine Learning combined with knowledge of neuroscience could also
positively affect situations in which humans may at times fail to make the best
decision. A similar situation is deciding which would be the best approach for
their child to learn. It is extremely tricky as the guardians have to make the
decision which may or may not be the best for the child. The Machine Learning
software could adopt various learning methods informed by developmental
psychology and measuring the reactions of the child to see which better suits
their aptitude. This way we can categorize and take account of emotional
responses. It can note various cues such as changes in the level of neurotransmitters
(as they can inform us certainly of their reactions as children cannot express
themselves thoroughly) such as dopamine to measure which method invokes
excitement which could nurture curiosity and a positive outlook on learning.
The field
of Artificial Neural Networks combines the best of my interests which is
programming, psychology and neuroscience. Through counselling and journaling, I
can record my thoughts and note how they have progressed to spot thinking loops.
These observations have helped me understand how I take decisions and what
factors influence me. My insights in psychology have formed through personal experience
and research through articles and books. It provides a unique perspective on the
way data is structured and processed for Machine Learning because I use these
logical structures to make decisions in my day-to-day life.
When I had
to select an educational path, I started noting which facilities were available
in my childhood to understand why I had developed certain hobbies. The search
for cause-and-effect models which could have influenced my temperament is
something I often deliberate and am proud to say that some hypotheses are still
under observation and experimentation.
Although I
try to learn more about my interests, my knowledge is scattered. I would love
to utilize the opportunity to take part in The Knowledge Society to structure
my learning.
It provides
me immense satisfaction that my daily experiences can be channeled to result in
user-friendly algorithms, that could be transferred to different scenarios.