The curse and blessing of dimensionality
Hi, I'm Hamza, a second-year Bachelor of Science in Artificial Intelligence student, currently studying in China — but my ambition is global. I'm not just studying AI; I'm living it, breathing it, and building with it. I blend the mindset of a developer, the soul of a designer, and the vision of a founder. I’ve built over 100+ real-world projects across AI, Cybersecurity, Web/App Development, and Full-Stack Engineering, including speech-to-text tools, intelligent systems, and creative front-end interfaces — not just as “practice,” but as purpose-driven solutions to real-life problems. One of my favorite builds? A speech-to-text tool I created during a hand injury — because for me, obstacles spark innovation. Beyond code, I’m a poet, artist, psychologist at heart, and a natural storyteller. I bring empathy into technology, and design into logic. My mission isn’t just to land jobs — it’s to create jobs, build ecosystems, and empower others through tech. I’ve been a speaker at AI conferences in China, an intern in design and web development, and I'm actively working toward launching a full suite of startups under "The Amber Origin" ecosystem — including AmberLingo, AmberNotes, Amber.ai, and more. 🌍 My Core Beliefs: Build with intention, not just completion. Learn fast, apply faster. Use tech to empower, not just impress. Creativity and code can co-exist beautifully. Let’s connect, collaborate, and maybe even create the next wave of impactful tech — together.
Introduction:
Today I did a deep research on one of the foundational topic in AI/ML and that is dimensionality. So here is my turn on what I understood from this topic as an second year BSAI student.
What exactly is dimensionality in AI?
it’s simply is a feature or variable in dataset.
A dataset with 1 feature —> 1D
2 feature —> 2D
1000 feature —> 1000D
So Dimensionality means number of input variable used to describe data points.
The curse of dimensionality
This term first coined by Richard Bellman. Basically the issue that arise with number of dimension increases this means
More Dimensions, The more issues
Problems
The volume of space increases exponentially.
Data points become far apart from each other.
High dimension —> data becomes sparse.
Analogy
Imagine trying to find your lost keys in a room (3D space). Easy. Now imagine searching in a 1000-dimensional room — suddenly everything is sparse, distances become meaningless, and your intuition fails.
Distance metrices break down:
Most ML algorithms (KNN, SVM, CLUSTERING) rely on distance.
In high-dimension, all distance start to look the same.
Hence models that depends on closeness perform poorly.
Exponential Computation:
Memory and computation needs to grow rapidly.
A grid search or brute force search becomes computationally intractable.
Overfitting:
More features = more flexibility = more risk of overfitting.
Model may memorize noise instead of patterns.
The blessing of dimensionality
In high dimension it is easy to separate data points with hyperplanes.
Many complex problems become linearly separable in high dimension.
Neural Network works well in high dimensional spaces.
More dimension = More ways to represent complex patterns.
The network can learn rich abstracts features because of dimensionality.
USES
KNN, CLUSTERING —> CURSE
SVM with KernelTrick —> BLESSING
Deep learning —> BLESSING
NLP —> BLESSING
