6 lessons we all need to learn from Machine Learning
Machine learning (ML) as a field has been around for few decades at-least. Thanks to deep learning, the term in now main-stream and no longer confined to research and academia.
Behind all the complex math of activation functions and back-propagation, deep learning is an abstraction of the network of neurons in our brain and simulate their fascinating capability to learn and generalize information.
The stupendous success of deep learning implies that there is value in devising algorithms inspired from humans. How about the contrary ?
Is there anything valuable for us to learn and be inspired from the practice of machine learning ?
Machine learning is more than neural networks and deep learning. It is a field with a legion of smart algorithms that deduce complex patterns and make predictions about the unknown.
Here are six things that I found to be very compelling in the field of ML and can potentially teach some life lessons
- Believe in the power of diversity (Random Forests)
The robustness of Random forests is contributed to its collection of distinct decision trees, each trying to solve part of the problem.
There is no forest with a single tree.
Similarly when people and groups from diverse backgrounds collaborate and work together, the results are heart warmingly good. If we confine ourselves only to our small communities that relates to us the most, we are limiting our true potential.
2. Learn from your mistakes, Always (Gradient Boosting)
Instead of using a collection of different classifiers, Gradient boosting leverages on the mistakes and continuously improve the base classifier in an iterative manner.
Like Gradient boosting machines, we need to realise that no one is perfect and we all make mistakes. It is important to identify the mistakes and learn the lessons that life teaches you.
Don’t get carried away by your accomplishments, rather focus on rectifying the defects at every stage of your life.
3. Every problem has a solution (Kernel methods & SVM)
Kernel methods are popular with support vector machines. They are used to transform the inseparable data in lower dimensions into higher dimensions and separate the classes using decision boundaries.
Life is not roses and rainbows all the time. It hits us with hard baffling problems to which there are no off-the-rack solutions. In those moments of despair, it is comforting to know that there are higher dimensions that you are not aware of and there is always a way out.
Stay calm and pray to the higher dimensions
4. Be receptive to alter personal beliefs (Bayes theorem)
Bayes theorem updates the prior likelihood of an event after acquiring additional information about related and conditional events.
Likewise, we all have preconceived beliefs and assumptions about people, cultures and society. These beliefs are cultivated based on our social, economic and demographic conditions.
It is critical that we consciously update ourselves when we experience new conditions that challenge our previous convictions.
Never be agnostic to changes.
5. Decisions should be made to account for the present (Gradient Descent)
From regression to neural networks, Gradient descent is the learning part of many Machine Learning algorithms. The principle is to always follow the path of steepest descent to reach the desired destination.
Often times we are faced with situations that require us to make a decision and choose one option among a myriad of alternatives. These decisions have a potential impact on the unknown future.
Gradient descent provides us a short-cut for such scenarios by always making the optimal choice for the present, not worrying too much about your future course.
Don’t wait for the ideal solution, you will never know what’s in store.
6. Relationships are subjected to change over time (k-means)
k-means algorithm iteratively improve the cluster quality by assigning and re-assigning data to different groups until an equilibrium is reached.
Human interactions and emotions play a crucial role in establishing our relationships. Not surprisingly, we get devastated when these relations are strained.
We should be aware that even strong relationships can change or end at any time for no obvious fault, just like in k-means. Accept the facts and cherish your loved ones when you can.
Let me know if you have similar analogies from your experience with algorithms and other learning paradigms. We can all learn from each other, including machine learning.
