Where should Burwood learners look for the best AI and programming classes?
To live online classes where a teacher takes Burwood teenagers and adults from Python into machine learning, building each model by hand before using a library. The 2021 Census counted 40,217 people in the Burwood Council area, with a median age of 34 and 12.5% aged 20 to 24. Lessons are one to one or in a small group at one level, and the first lesson is free.
An autoencoder is a neural network asked to copy its input through a narrow middle layer. If it can rebuild the input from only a few numbers, those numbers have captured what matters. We gave one the monthly entry pattern of every NSW train station with a full record, 270 of them over 23 months, and a middle layer of just two numbers. Trained by gradient descent from random weights, its error fell from 0.011495 to 0.006155, exactly the figure principal component analysis gives directly. Burwood's pattern came back with an error of 0.0300, closer than most, because its year looks a lot like the network's average year.
Facts last verified 10 October 2026. Teaching is online; no Burwood branch is claimed.
Each course builds the skills the autoencoder project uses. A free trial lesson sets the level, with no payment details taken.

Machine learning from scratch in Python: regression, classification and small neural networks.
See the syllabus
AI from foundations to deep learning, with the maths explained as it is used.
See the syllabus
Data handling, dimensionality reduction and modelling on real public datasets.
See the syllabus
The Python needed before any of the AI work, taught properly from the start.
See the syllabusBrowse the course atlas for more than one hundred options and use the coding roadmap to check prerequisites.
The four we are known for
These run underneath everything above. Every one is live and online, placed by ability rather than by age, and the first class is free.

Python, web and AI projects where the learner still owns the thinking.
See the syllabus
Automate the work you already do, then let AI carry part of it.
See the syllabus
Train a model, read what it learned, and be able to say why it is wrong.
See the syllabus
Run AI coding agents on real work without losing control of the codebase.
See the syllabusFrom the 2021 Census QuickStats for the Burwood local government area, compared with New South Wales and Australia.
| Item | Burwood |
|---|---|
| Population | 40,217 |
| Private dwellings | 16,160 |
| Median age | 34 |
| Mean household size | 2.7 |
| Share | Burwood | NSW | Australia |
|---|---|---|---|
| People aged 20 to 24 | 12.5% (5,006) | 6.1% | 6.2% |
| Flats or apartments, occupied homes | 49.1% (6,944) | 21.7% | 14.2% |
| Bachelor degree or higher, aged 15 and over | 37.8% (13,395) | 27.8% | 26.3% |
| Professionals, employed residents | 30.2% (5,659) | 25.8% | 24.0% |
One Burwood resident in eight was aged 20 to 24, roughly double the proportion across New South Wales.
Flats or apartments made up 49.1% of occupied homes, against 34.5% that were separate houses.
Homes held 2.7 people on average, slightly above the NSW figure of 2.6.
To keep individuals unidentifiable, the bureau perturbs each cell slightly, which means component figures need not add up to the published total. We report published numbers only and calculate no totals of our own.
Each station's 23 monthly entry counts are divided by its own average, so a busy station and a quiet one can be compared by shape alone. The network's average shape is then removed, leaving only what makes each station different.
| Numbers kept per station | Burwood error | Median error | Burwood rank, 1 = least typical |
|---|---|---|---|
| 0 (average shape only) | 0.0355 | 0.0587 | 257 of 270 |
| 1 | 0.0309 | 0.0504 | 227 of 270 |
| 2 | 0.0300 | 0.0425 | 192 of 270 |
| 3 | 0.0277 | 0.0400 | 191 of 270 |
| Training step | Mean squared error |
|---|---|
| Random start | 0.011495 |
| After 1,000 epochs | 0.006284 |
| After 30,000 epochs | 0.006155 |
| Principal component analysis, no training | 0.006155 |
With nothing but the network's average shape, Burwood's error is 0.0355 against a median of 0.0587. Only 13 stations sit closer to the average year.
Each extra number shrinks the typical station's error faster than Burwood's, so Burwood slides from 257th to 192nd as the model learns the common ways stations differ.
A linear autoencoder trained by gradient descent ends at 0.006155, the error PCA reaches in one calculation. With no nonlinearity, the network can only rediscover PCA.
Bankstown is among the three hardest stations to rebuild at every bottleneck size, and Clarendon at three of the four. A large error flags an unusual pattern of months; it does not reveal why, and we make no claim about the cause for any station.
The profiles, components, errors, ranks and training run are our own calculations from the Transport for NSW file, with 26 train stations left out because at least one month was missing or published only as "Less than 50". The weights start from a fixed random seed so that learners can reproduce every number.
Read directly from the published monthly entries for Burwood Station.
| Month or measure | Entries |
|---|---|
| Lowest month, January 2025 | 418,631 |
| Highest month, March 2026 | 563,430 |
| August 2026 | 553,354 |
| Monthly mean over 23 months (our calculation) | 502,685 |
Divided by its mean, Burwood's January 2025 comes to 0.833, while the average station's January 2025 stands at 0.812. That small difference is typical of Burwood: its dips and peaks follow the network closely, which is exactly why the autoencoder rebuilds it well.
Normalise the profiles with NumPy, compute the average shape, and find which stations differ most from it.
Write the autoencoder's forward pass and gradients by hand, train it, and check the loss against PCA.
Use reconstruction error to flag unusual records in work data, and learn when a nonlinear autoencoder is worth it.
Transport for NSW, the NSW Department of Education, Burwood Council and the statistics bureau publish this open data and are not associated with Modern Age Coders. The modelling belongs to us.
From the NSW schools file, the station counts and the 2027 school calendar.
Its first teacher is recorded in April 1858, the earliest date among the government schools of the Burwood Council area.
Croydon Public School follows in 1884 and Croydon Park Public School in 1886, with Enfield Public School in 1924.
Croydon Station counted 58,449 entries in August 2026, about a tenth of Burwood Station's 553,354.
School year 2027
NSW Eastern division Term 3 runs from Tuesday 20 July to Friday 24 September 2027, and Term 4 from Tuesday 12 October to Monday 20 December. Group lessons follow those breaks.
The age bands are only a guide; the trial lesson decides.
Scratch and puzzles that build sequencing, patterns and debugging.
Scratch Coding for KidsProblem Solving and Computational Thinking for KidsPython done properly, then the maths that machine learning leans on.
Python for TeensMaths Through CodingModels built from first principles and tested honestly.
AI and Machine Learning for TeensData Science for TeensNeural networks, dimensionality reduction and full data projects.
Artificial Intelligence CourseData Science CourseBecause knowing what a network can and cannot learn is what stops you trusting one that has learned nothing new.
A library will train an autoencoder in three lines. Only someone who has written the gradients knows that, without a nonlinear layer, thirty thousand epochs of training end exactly where PCA started, at 0.006155. That is a lesson about architecture no amount of calling functions teaches.
Learners who have watched Burwood's error fall from 0.0355 to 0.0300 also learn to read reconstruction error with care. It can point at unusual records, but it says nothing about causes, and they will not let a model pretend otherwise.
In 2026 anyone can import a neural network, and the Burwood learner who knows when it is secretly a straight line is the one who designs one that is not. The longer argument is in why teenagers should still learn to code in 2026.
Schools in the council area have addresses in Burwood, Croydon, Croydon Park and Enfield; an online class serves each equally.
Learner and teacher work in the same notebook, so every line of a model is written and explained live.
Topics match the NSW stages, and lessons break when the state school holidays begin.
The trial is a full lesson, finishing with a recommended course, level and time and no charge.
Fifteen to twenty learners at one level per group, or one to one if preferred.
Two set lessons each week, around eight per month.
A 7pm Burwood lesson in daylight saving time begins at 1:30pm India time; in winter, at 2:30pm there.
Sydney time, always
Daylight saving in Sydney begins on Sunday 4 October 2026 and ends on Sunday 4 April 2027, while India stays on one time all year. Lesson times are given in Sydney time.
A free start and two monthly options.
The opening lesson is free and ends with a plan.
Group format: around eight lessons a month among fifteen to twenty learners.
One-to-one format: around eight lessons a month with your own teacher.
Learners outside India pay in US dollars, so Burwood fees match everywhere else and there is no area-specific list. You pay only after the trial, once the course and time are decided; holidays, missed lessons and changes of format are set out on the pricing page.
Rated 4.9 across 547 Google reviews. These are real reviews, reproduced as written.
★★★★★
"The one step solution for my son. Modern Age Coders make learning coding so simple that kids love it. The teachers explain complex concepts clearly with practical exercises and interactive content."
Ria Mukherjee
Parent
★★★★★
"Modern Age Coders has been a game-changer for me. I struggled to grasp IT concepts and coding before joining, but their classes transformed everything. I can now confidently write complex programs with ease."
Samriddha Mondal
Student
★★★★★
"One of the most wonderful education centres out there. Education is not limited to school syllabus but focuses on skill development."
Vansh Agarwal
Student
★★★★★
"My child Dhairya is really enjoying the Modern Age Coders classes. This is his first online class and he eagerly looks forward to it. I can already see his improvement, and the teachers are very cooperative."
Sonam Oswal
Parent of Dhairya
★★★★★
"Modern Age Coders have wonderful teachers who teach in a clear, easy and practical way. The teacher boosts students' confidence and inspires them to learn without hesitation."
Sonu Goyal
Parent
★★★★★
"I highly recommend this computer coding class! The teachers are incredibly knowledgeable and passionate about coding."
Ritu Kedia
Parent
Share the learner's age and interests. A Year 11 student could begin by normalising station profiles; an adult could start by training the autoencoder.
What Burwood families and adult learners ask most often.
QuickStats from the 2021 Census give 40,217 people and 16,160 private dwellings for the Burwood local government area.
An autoencoder is a neural network trained to reproduce its input after passing it through a narrow middle layer. The middle layer becomes a compressed summary, and the reconstruction error shows how well the summary captures the input.
Learners compress the 23-month entry pattern of 270 NSW train stations into two numbers each, rebuild the patterns, and measure the error. Burwood Station's error is 0.0300 against a median of 0.0425.
A linear autoencoder with squared error can only learn the same subspace principal component analysis finds. Gradient descent reached 0.006155, the PCA error, after enough epochs.
Not necessarily. It means the station's pattern across months is unusual compared with the rest. The data does not say why, and the project draws no conclusions about causes.
Teenagers need confident algebra and some practice with Python lists. Matrix ideas are taught as the project needs them.
Yes. From age six, children start with Scratch and problem-solving, and move to Python when ready.
No. All lessons happen live online, and we run no premises in Australia. A computer with audio and a dependable connection is all that is needed; the number we list is in India.
The first lesson is free. After that a group place is USD 100 a month for two live lessons a week, roughly eight lessons, with fifteen to twenty learners, or USD 150 a month one to one on the same timetable. Course, format and time are fixed before any payment.
By level, pace and goals, in groups of fifteen to twenty, regardless of age or location. One-to-one lessons are available when no group time works.
Neighbouring Strathfield has a page on percentile ranks, and Ku-ring-gai one on lasso regression. For the wider picture, see the Sydney coding classes page or classes across New South Wales.