About Adam Optimizer Visualizer
This Adam optimizer visualizer runs Adam on the Rosenbrock function at a learning rate of 0.05, with SGD, momentum and RMSprop for comparison. It is free, needs no sign-up and runs in your browser, so nothing you enter is uploaded.
The Rosenbrock function has a long, curved, almost flat valley leading to the minimum at (1, 1), and walls that are very steep. Plain gradient descent must use a tiny learning rate to survive the walls, then barely moves along the floor. Adam divides each step by the size of the recent gradients, so it takes useful steps on the flat floor without overshooting the walls.
Adam also keeps a momentum term and corrects the bias of its running averages at the start, which is why its first steps are the right size. Press Play and compare how far each optimizer has got.
How to use Adam Optimizer Visualizer
- Choose a function. The suggested learning rate for it is filled in.
- Change the learning rate if you like. Smaller is slower and safer, larger is faster until it blows up.
- Press Step to move every optimizer once, or Play to animate. "+25 steps" skips ahead.
- Tap the plot to move the starting point (the cross). The paths restart from there.
- Read the table for each optimizer's position, value and status: running, converged or diverged.
- Watch the log chart: a line that keeps falling is converging, and a line that climbs is diverging.
Features
- Seven functions: three curves and four 2D surfaces, including a narrow valley, Rosenbrock, Himmelblau and a saddle.
- SGD, Momentum, RMSprop and Adam, all from the same start with the same learning rate.
- Contour map for 2D functions and a curve for 1D, with a coloured path for each optimizer.
- Divergence detection that stops a run and says which optimizers blew up.
- A log-scale chart of the function value against step.
- Step, play and +25 steps controls, with slower playback if you prefer reduced motion.
- Tap the plot to choose the starting point.
Adam update
m and s are running averages of the gradient and of its square; t counts the steps.
m = 0.9 m + 0.1 g
s = 0.999 s + 0.001 g^2
x = x - lr (m / (1 - 0.9^t)) / (sqrt(s / (1 - 0.999^t)) + eps)Worked examples
Rosenbrock, learning rate 0.05, 300 steps gives SGD diverged, Momentum diverged, RMSprop 3.8992, Adam 1.7147
A learning rate this large diverges for SGD and Momentum, because the walls are steep. RMSprop and Adam survive it and keep descending, though neither has reached the minimum yet.
Privacy: how it works
- Everything is computed in your browser tab. Nothing is sent to Ulkaa or anyone else.
Tips
- On the parabola, try a learning rate of 0.5: SGD jumps to the minimum in a single step because the step size matches the curvature exactly.
- On the double well, drag the start to the left side and see it settle in the other, shallower minimum.
- On Himmelblau, change the start point and see different optimizers choose different minima.
- If a path vanishes off the plot, check the status column: it probably diverged.
Frequently asked questions
Is it really free?
Yes. There is no price, no trial and no credit card. Ulkaa Tool does the work with your own device, so there is no server bill to pass on to you.
Do I need an account?
No. There is no sign-up and no login. Open the page and use it. An Ulkaa account is not needed for anything on this tool.
Does it run in my browser?
Yes. Your settings are processed by code running in your browser tab, on your own device.
Is my data uploaded?
No. Your settings are never sent to Ulkaa or to anyone else. You can check this yourself by watching the Network tab in your browser developer tools while you use the tool.
How does Adam work?
It keeps a running average of the gradient (momentum) and of the squared gradient, corrects their start-up bias, and scales each step by the ratio of the two.
Why is Adam popular?
Its defaults work across many problems with little tuning, and it handles directions with very different gradient sizes.
More Gradient Descent Visualizer pages
- Gradient Descent VisualizerCompare SGD, Momentum, RMSprop and Adam side by side, step by step, on curves and contour maps.
- Learning Rate Too High: Gradient Descent DivergenceSee the error grow instead of shrink when the learning rate crosses its limit.
- SGD Learning Rate VisualizerChange the learning rate and watch plain gradient descent speed up, oscillate or stall.
- Momentum vs SGDSee how a running velocity helps gradient descent in a long, narrow valley.
- RMSprop Optimizer VisualizerSee RMSprop normalise its steps on a surface with four minima.
- Gradient Descent and Local MinimaStart on one side of a double well and see which minimum each optimizer reaches.
- Gradient Descent at a Saddle PointStart almost on a ridge and see how long each optimizer takes to escape the saddle.
- Gradient Descent on the Rosenbrock FunctionFollow gradient descent down the curved banana valley of the Rosenbrock function.
- Gradient Descent on a Bumpy LossSee how a bumpy loss traps plain gradient descent and how momentum helps.