Too much math

Thread Starter

Yuseph

Joined Jun 8, 2020
45
Dont worry guys i didnt give up on electronics. On the contrary. I had to start math from scratch I have another month of work before reaching 12th grade scientific class level. I assumed thats what it takes to tackle machine learning maths.
 

dl324

Joined Mar 30, 2015
11,885
I don't know what you mean by "machine learning math", but electronics requires more than just high school math. You'll need good algebra skills for calculus. Trig would also help, but I never took it in high school or college. I picked up what I needed to know about trig when I was learning calculus.
 

bogosort

Joined Sep 24, 2011
519
You'll need good algebra skills for calculus. Trig would also help, but I never took it in high school or college. I picked up what I needed to know about trig when I was learning calculus.
I did the opposite. Due to a clerical blunder, I was able to graduate high school without algebra. In college, I picked up the basics while learning trig and calculus, and then became proficient in a differential equations course.

My vote for "machine learning math" is linear algebra.
 

dl324

Joined Mar 30, 2015
11,885
My vote for "machine learning math" is linear algebra.
I never did take linear algebra. Didn't know what I was missing until I needed to analyze/modify some graphical data and I needed to learn some basic linear algebra to be able to do coordinate transformations as cell instances were rotated, flipped, and/or placed at different levels of hierarchy.
 

dl324

Joined Mar 30, 2015
11,885
The trick to math is repetition until it becomes intuitively automatic.
You also lose it if you don't use it. Most of the electronics I do these days only requires algebra and arithmetic. Anything else and I just look up the solutions for integrals, plug in values, and do the arithmetic.
 

tindel

Joined Sep 16, 2012
734
In engineering, you don't generally have to remember how to do something perfectly... you have to know the concepts. Limits, differentials, integrals... very rarely do you need to solve any of these things by hand, but you may need to be able to look at a graph and note a non-linearity, the slope of a line (derivative), how integrals of the time domain can be used to evaluate the same signal in the frequency domain, how discrete math can then also simplify that integral to come up with algorithms that can quickly calculate the FFT on you scope, etc. These are things where you don't need to generally recall the specifics, but you do need to understand the concepts to be successful.

Hell, capacitors integrate and inductors differentiate.... EEs do calculus every day and don't always think about it.

If you're working in R&D it's different though. Many new discoveries depend on a through knowledge of calculus, linear algebra, discrete math, geometry, trig, physics, etc. That's why general knowledge of these topics are required for a bachelors of science. A master of science degree requires some application of these topics. A doctorate requires in-depth knowledge of these things to provide novel contributions to a field of study.
 

sagor

Joined Mar 10, 2019
283
Let's not forget Fourier Analysis (Transform) and Laplace transformations. Never did use those after learning them. Forgot about them except for their names.
 

Thread Starter

Yuseph

Joined Jun 8, 2020
45
Let's not forget Fourier Analysis (Transform) and Laplace transformations. Never did use those after learning them. Forgot about them except for their names.
Dude, what you talking about. I was eager to attain laplace transforms fourier series and z-transforms only to show off and you're telling me i'll never have an opportunity to use them ?
 

tindel

Joined Sep 16, 2012
734
I use transforms all of the time... nearly daily... A good engineer can jump from time to frequency domain and back easily. Do I solve the actual transform? Sometimes... not often. But I know an RC time constant has a -3dB point at 1/(2*pi*R*C) in the f-domain. If the -3dB point is at 100Hz I know if I drive the circuit with a PWM at 1kHz it will be filtered by -20dB... therefore, on a 3.3V PWM I expect 330mVac to couple to the output of the PWM. I can then decide if that's good enough - or not.

I also use it to determine stability... of opamps, power supplies, and systems...

I literally use that knowledge almost every day in some form or fashion. Rarely do I pull out a pencil or pen to do the analysis, though, I can, and do occasionally.
 

cmartinez

Joined Jan 17, 2007
7,174
Dude, what you talking about. I was eager to attain laplace transforms fourier series and z-transforms only to show off and you're telling me i'll never have an opportunity to use them ?
Yes you will definitely have an opportunity to use them, if you choose the right career path...
 

Wolframore

Joined Jan 21, 2019
1,962
I use transforms all of the time... nearly daily... A good engineer can jump from time to frequency domain and back easily. Do I solve the actual transform? Sometimes... not often. But I know an RC time constant has a -3dB point at 1/(2*pi*R*C) in the f-domain. If the -3dB point is at 100Hz I know if I drive the circuit with a PWM at 1kHz it will be filtered by -20dB... therefore, on a 3.3V PWM I expect 330mVac to couple to the output of the PWM. I can then decide if that's good enough - or not.

I also use it to determine stability... of opamps, power supplies, and systems...

I literally use that knowledge almost every day in some form or fashion. Rarely do I pull out a pencil or pen to do the analysis, though, I can, and do occasionally.
I agree but it depends on what area you’re in.

The phase of -3 dB point is 45 degree and cosecant of 0.707... the transform forms the basis of our complex analysis. Once you have the understanding its possible to just see the relationships. All our formulas for complex analysis is derived from this and people work through omega whether they know it or not.

Perhaps you will be quizzed and asked to solve at an interview.
 
Last edited:
Top