ans = 
Results for
When solving Cody problems, sometimes your solution takes too long — especially if you’re recomputing large arrays or iterative sequences every time your function is called.
The Cody work area resets between separate runs of your code, but within one Cody test suite, your function may be called multiple times in a single session.
This is where persistent variables come in handy.
A persistent variable keeps its value between function calls, but only while MATLAB is still running your function suite.
This means:
- You can cache results to avoid recomputation.
- You can accumulate data across multiple calls.
- But it resets when Cody or MATLAB restarts.
Suppose you’re asked to find the n-th Fibonacci number efficiently — Cody may time out if you use recursion naively. Here’s how to use persistent to store computed values:
function f = fibPersistent(n)
import java.math.BigInteger
persistent F
if isempty(F)
F=[BigInteger('0'),BigInteger('1')];
for k=3:10000
F(k)=F(k-1).add(F(k-2));
end
end
% Extend the stored sequence only if needed
while length(F) <= n
F(end+1)=F(end).add(F(end-1));
end
f = char(F(n+1).toString); % since F(1) is really F(0)
end
%calling function 100 times
K=arrayfun(@(x)fibPersistent(x),randi(10000,1,100),'UniformOutput',false);
K(100)
The fzero function can handle extremely messy equations — even those mixing exponentials, trigonometric, and logarithmic terms — provided the function is continuous near the root and you give a reasonable starting point or interval.
It’s ideal for cases like:
- Solving energy balance equations
- Finding intersection points of nonlinear models
- Determining parameters from experimental data
Example: Solving for Equilibrium Temperature in a Heat Radiation-Conduction Model
Suppose a spacecraft component exchanges heat via conduction and radiation with its environment. At steady state, the power generated internally equals the heat lost:
Given constants:
= 25 W- k = 0.5 W/K
- ϵ = 0.8
- σ = 5.67e−8 W/m²K⁴
- A = 0.1 m²
= 250 K
Find the steady-state temperature, T.
% Given constants
Qgen = 25;
k = 0.5;
eps = 0.8;
sigma = 5.67e-8;
A = 0.1;
Tinf = 250;
% Define the energy balance equation (set equal to zero)
f = @(T) Qgen - (k*(T - Tinf) + eps*sigma*A*(T.^4 - Tinf^4));
% Plot for a sense of where the root lies before implementing
fplot(f, [250 300]); grid on
xlabel('Temperature (K)'); ylabel('f(T)')
title('Energy Balance: Root corresponds to steady-state temperature')
% Use fzero with an interval that brackets the root
T_eq = fzero(f, [250 300]);
fprintf('Steady-state temperature: %.2f K\n', T_eq);
I set my 3D matrix up with the players in the 3rd dimension. I set up the matrix with: 1) player does not hold the card (-1), player holds the card (1), and unknown holding the card (0). I moved through the turns (-1 and 1) that are fixed first. Then cycled through the conditional turns (0) while checking the cards of each player using the hints provided until it was solved. The key for me in solving several of the tests (11, 17, and 19) was looking at the 1's and 0's being held by each player.
sum(cardState==1,3);%any zeros in this 2D matrix indicate possible cards in the solution
sum(cardState==0,3)>0;%the ones in this 2D matrix indicate the only unknown positions
sum(cardState==1,3)|sum(cardState==0,3)>0;%oring the two together could provide valuable information
Some MATLAB Cody problems prohibit loops (for, while) or conditionals (if, switch, while), forcing creative solutions.
One elegant trick is to use nested functions and recursion to achieve the same logic — while staying within the rules.
Example: Recursive Summation Without Loops or Conditionals
Suppose loops and conditionals are banned, but you need to compute the sum of numbers from 1 to n. This is a simple example and obvisously n*(n+1)/2 would be preferred.
function s = sumRecursive(n)
zero=@(x)0;
s = helper(n); % call nested recursive function
function out = helper(k)
L={zero,@helper};
out = k+L{(k>0)+1}(k-1);
end
end
sumRecursive(10)
- The helper function calls itself until the base case is reached.
- Logical indexing into a cell array (k>0) act as an 'if' replacement.
- MATLAB allows nested functions to share variables and functions (zero), so you can keep state across calls.
Tips:
- Replace 'if' with logical indexing into a cell array.
- Replace for/while with recursion.
- Nested functions are local and can access outer variables, avoiding global state.
Many MATLAB Cody problems involve recognizing integer sequences.
If a sequence looks familiar but you can’t quite place it, the On-Line Encyclopedia of Integer Sequences (OEIS) can be your best friend.
OEIS will often identify the sequence, provide a formula, recurrence relation, or even direct MATLAB-compatible pseudocode.
Example: Recognizing a Cody Sequence
Suppose you encounter this sequence in a Cody problem:
1, 1, 2, 3, 5, 8, 13, 21, ...
Entering it on OEIS yields A000045 – The Fibonacci Numbers, defined by:
F(n) = F(n-1) + F(n-2), with F(1)=1, F(2)=1
You can then directly implement it in MATLAB:
function F = fibSeq(n)
F = zeros(1,n);
F(1:2) = 1;
for k = 3:n
F(k) = F(k-1) + F(k-2);
end
end
fibSeq(15)
When solving MATLAB Cody problems involving very large integers (e.g., factorials, Fibonacci numbers, or modular arithmetic), you might exceed MATLAB’s built-in numeric limits.
To overcome this, you can use Java’s java.math.BigInteger directly within MATLAB — it’s fast, exact, and often accepted by Cody if you convert the final result to a numeric or string form.
Below is an example of using it to find large factorials.
function s = bigFactorial(n)
import java.math.BigInteger
f = BigInteger('1');
for k = 2:n
f = f.multiply(BigInteger(num2str(k)));
end
s = char(f.toString); % Return as string to avoid overflow
end
bigFactorial(100)
From my experience, MATLAB's Deep Learning Toolbox is quite user-friendly, but it still falls short of libraries like PyTorch in many respects. Most users tend to choose PyTorch because of its flexibility, efficiency, and rich support for many mathematical operators. In recent years, the number of dlarray-compatible mathematical functions added to the toolbox has been very limited, which makes it difficult to experiment with many custom networks. For example, svd is currently not supported for dlarray inputs.
This link (List of Functions with dlarray Support - MATLAB & Simulink) lists all functions that support dlarray as of R2026a — only around 200 functions (including toolbox-specific ones). I would like to see support for many more fundamental mathematical functions so that users have greater freedom when building and researching custom models. For context, the core MATLAB mathematics module contains roughly 600 functions, and many application domains build on that foundation.
I hope MathWorks will prioritize and accelerate expanding dlarray support for basic math functions. Doing so would significantly increase the Deep Learning Toolbox's utility and appeal for researchers and practitioners.
Thank you.
Run MATLAB using AI applications by leveraging MCP. This MCP server for MATLAB supports a wide range of coding agents like Claude Code and Visual Studio Code.
Check it out and share your experiences below. Have fun!
GitHub repo: https://github.com/matlab/matlab-mcp-core-server
Yann Debray's blog post: https://blogs.mathworks.com/deep-learning/2025/11/03/releasing-the-matlab-mcp-core-server-on-github/
Hey Relentless Coders! 😎
Let’s get to know each other. Drop a quick intro below and meet your teammates! This is your chance to meet teammates, find coding buddies, and build connections that make the contest more fun and rewarding!
You can share:
- Your name or nickname
- Where you’re from
- Your favorite coding topic or language
- What you’re most excited about in the contest
Let’s make Team Relentless Coders an awesome community—jump in and say hi! 🚀
Welcome to the Cody Contest 2025 and the Relentless Coders team channel! 🎉
You never give up. When a problem gets tough, you dig in deeper. This is your space to connect with like-minded coders, share insights, and help your team win. To make sure everyone has a great experience, please keep these tips in mind:
- Follow the Community Guidelines: Take a moment to review our community standards. Posts that don’t follow these guidelines may be flagged by moderators or community members.
- Ask Questions About Cody Problems: When asking for help, show your work! Include your code, error messages, and any details needed to reproduce your results. This helps others provide useful, targeted answers.
- Share Tips & Tricks: Knowledge sharing is key to success. When posting tips or solutions, explain how and why your approach works so others can learn your problem-solving methods.
- Provide Feedback: We value your feedback! Use this channel to report issues or share creative ideas to make the contest even better.
Have fun and enjoy the challenge! We hope you’ll learn new MATLAB skills, make great connections, and win amazing prizes! 🚀
I'm working on training neural networks without backpropagation / automatic differentiation, using locally derived analytic forms of update rules. Given that this allows a direct formula to be derived for the update rule, it removes alot of the overhead that is otherwise required from automatic differentiation.
However, matlab's functionalities for neural networks are currently solely based around backpropagation and automatic differentiation, such as the dlgradient function and requiring everything to be dlarrays during training.
I have two main requests, specifically for functions that perform a single operation within a single layer of a neural network, such as "dlconv", "fullyconnect", "maxpool", "avgpool", "relu", etc:
- these functions should also allow normal gpuArray data instead of requiring everything to be dlarrays.
- these functions are currently designed to only perform the forward pass. I request that these also be designed to perform the backward pass if user requests. There can be another input user flag that can be "forward" (default) or "backward", and then the function should have all the necessary inputs to perform that operation (e.g. for "avgpool" forward pass it only needs the avgpool input data and the avgpool parameters, but for the "avgpool" backward pass it needs the deriviative w.r.t. the avgpool output data, the avgpool parameters, and the original data dimensions). I know that there is a maxunpool function that achieves this for maxpool, but it has significant issues when trying to use it this way instead of by backpropagation in a dlgradient type layer, see (https://www.mathworks.com/matlabcentral/answers/2179587-making-a-custom-way-to-train-cnns-and-i-am-noticing-that-avgpool-is-significantly-faster-than-maxpo?s_tid=srchtitle).
I don't know how many people would benefit from this feature, and someone could always spend their time creating these functionalities themselves by matlab scripts, cuDNN mex, etc., but regardless it would be nice for matlab to have this allowable for more customizable neural net training.
Edit 15 Oct 2025: Removed incorrect code. Replaced symmatrix2sym and symfunmatrix2symfun with sym and symfun respectively (latter supported as of 2024b).
The Symbolic Math Toolbox does not have its own dot and and cross functions. That's o.k. (maybe) for garden variety vectors of sym objects where those operations get shipped off to the base Matlab functions
x = sym('x',[3,1]); y = sym('y',[3,1]);
which dot(x,y)
dot(x,y)
which cross(x,y)
cross(x,y)
clearvars
x = symmatrix('x',[3,1]); y = symmatrix('y',[3,1]);
z = symmatrix('z',[1,1]);
sympref('AbbreviateOutput',false);
dot() expands the result, which isn't really desirable for exposition.
eqn = z == dot(x,y)
Also, dot() returns the the result in terms of the conjugate of x, which can't be simplifed away at the symmatrix level
assumeAlso(sym(x),'real')
class(eqn)
try
eqn = z == simplify(dot(x,y))
catch ME
ME.message
end
To get rid of the conjugate, we have to resort to sym
eqn = simplify(sym(eqn))
but again we are in expanded form, which defeats the purpose of symmatrix (et. al.)
But at least we can do this to get a nice equation
eqn = z == x.'*y
dot errors with symfunmatrix inputs
clearvars
syms t real
x = symfunmatrix('x(t)',t,[3,1]); y = symfunmatrix('y(t)',t,[3,1]);
try
dot(x,y)
catch ME
ME.message
end
Cross works (accidentally IMO) with symmatrix, but expands the result, which isn't really desirable for exposition
clearvars
x = symmatrix('x',[3,1]); y = symmatrix('y',[3,1]);
z = symmatrix('z',[3,1]);
eqn = z == cross(x,y)
And it doesn't work at all if an input is a symfunmatrix
syms t
w = symfunmatrix('w(t)',t,[3,1]);
try
eqn = z == cross(x,w);
catch ME
ME.message
end
In the latter case we can expand with
eqn = z == cross(sym(x),symfun(w)) % x has to be converted
But we can't do the same with dot (as shown above, dot doesn't like symfun inputs)
try
eqn = z == dot(sym(x),symfun(w))
catch ME
ME.message
end
Looks like the only choice for dot with symfunmatrix is to write it by hand at the matrix level
x.'*w
or at the sym/symfun level
sym(x).'*symfun(w) % assuming x is real
Ideally, I'd like to see dot and cross implemented for symmatrix and symfunmatrix types where neither function would evaluate, i.e., expand, until both arguments are subs-ed with sym or symfun objects of appropriate dimension.
Also, it would be nice if symmatrix could be assumed to be real. Is there a reason why being able to do so wouldn't make sense?
try
assume(x,'real')
catch ME
ME.message
end
Automating Parameter Identifiability Analysis in SimBiology
Is it possible to develop a MATLAB Live Script that automates a series of SimBiology model fits to obtain likelihood profiles? The goal is to fit a kinetic model to experimental data while systematically fixing the value of one kinetic constant (e.g., k1) and leaving the others unrestricted.
The script would perform the following:
Use a pre-configured SimBiology project where the best fit to the experimental data has already been established (including dependent/independent variables, covariates, the error model, and optimization settings).
Iterate over a defined sequence of fixed values for a chosen parameter.
For each fixed value, run the estimation to optimize the remaining parameters.
Record the resulting Sum of Squared Errors (SSE) for each run.
The final output would be a likelihood profile—a plot of SSE versus the fixed parameter value (e.g., k1)—to assess the practical identifiability of each model parameter.
Something that I periodically wonder about is whether an integration with the Rubi integration rules package would improve symbolic integration in Matlab's Symbolic Toolbox. The project is open-source with an MIT-licensed, has a Mathematica implementation, and supposedly SymPy is working on an implementation. Much of my intrigue comes from this 2022 report that compared the previous version of Rubi (4.16.1) against various CAS systems, including Matlab 2021a (Mupad):
While not really an official metric for Rubi, this does "feel" similar to my experience computing symbolic integrals in Matlab Symbolic Toolbox vs Maple/Mathematica. What do y'all think?
Have you ever been enrolled in a course that uses an LMS and there is an assignment that invovles posting a question to, or answering a question in, a discussion group? This discussion group is meant to simulate that experience.
The functionality would allow report generation straight from live scripts that could be shared without exposing the code. This could be useful for cases where the recipient of the report only cares about the results and not the code details, or when the methodology is part of a company know how, e.g. Engineering services companies.
In order for it to be practical for use it would also require that variable values could be inserted into the text blocks, e.g. #var_name# would insert the value of the variable "var_name" and possibly selecting which code blocks to be hidden.
For the www, uk, and in domains,a generative search answer is available for Help Center searches. Please let us know if you get good or bad results for your searches. Some have pointed out that it is not available in non-english domains. You can switch your country setting to try it out. You can also ask questions in different languages and ask for the response in a different language. I get better results when I ask more specific queries. How is it working for you?
Modern engineering requires both robust hardware and powerful simulation tools. MATLAB and Simulink are widely used for data analysis, control design, and embedded system development. At the same time, Kasuo offers a wide range of components—from sensors and connectors to circuit protection devices—that engineers rely on to build real-world systems.
By combining these tools, developers can bridge the gap between simulation and implementation, ensuring their designs are reliable and ready for deployment.
Example Use Case: Sensor Data Acquisition and Processing
- Kasuo Hardware Setup
- Select a Kasuo sensor (e.g., temperature, microphone, or motion sensor).
- Connect it to a DAQ or microcontroller board for data collection.
- Data Acquisition in MATLAB
- Use MATLAB’s Data Acquisition Toolbox to stream sensor data directly.
- Example snippet:
s = daq("ni");
addinput(s,
"Dev1", "ai0", "Voltage");
data = read(s, seconds(
5), "OutputFormat", "Matrix");
plot(data);
- Signal Processing with Simulink
- Build a Simulink model to filter noise, detect anomalies, or design control logic.
- Simulink enables real-time visualization and iterative tuning.
- Validation & Protection Simulation
- Add Kasuo’s circuit protection components (e.g., TVS diodes, surge suppressors) in the physical design.
- Use Simulink to simulate stress conditions, validating system robustness before hardware testing.
Benefits of the Workflow
- Faster prototyping with MATLAB & Simulink.
- Greater reliability by incorporating Kasuo protection devices.
- Seamless transition from model to hardware implementation.
Conclusion
Kasuo’s electronic components provide the hardware foundation for many embedded and signal processing applications. When combined with MATLAB and Simulink, engineers can design, simulate, and validate systems more efficiently—reducing risks and development time.
With AI agents dev coding on other languages has become so easy.
Im waiting for matlab to build something like warp but for matlab.
I know they have the current ai but with all respect it's rubbish compared to vibe coding tools in others sectors.
Matlab leads AI so it really should be leading this space.
Hello MATLAB Central community,
My name is Yann. And I love MATLAB. I also love Python ... 🐍 (I know, not the place for that).
I recently decided to go down the rabbit hole of AI. So I started benchmarking deep learning frameworks on basic examples. Here is a recording of my experiment:
Happy to engage in the debate. What do you think?
