About Gradient Descent and Local Minima
This page uses the double-well function x to the fourth minus 3x squared plus x, which has two minima of different depth, starting on the right-hand side at x = 1.9. It is free, needs no sign-up and runs in your browser, so nothing you enter is uploaded.
Gradient descent always goes downhill from where it starts, so it ends in the minimum whose basin it began in. The right-hand well here is shallower than the left, so a start on the right settles in the worse minimum even though a better one exists.
Tap the plot to start on the left and see the difference. With a larger learning rate and momentum, an optimizer may carry enough speed to cross the hill between the wells. This is one reason why training neural networks uses noise and momentum.
How to use Gradient Descent and Local Minima
- 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.
Worked examples
Double well, start 1.9, learning rate 0.02, 200 steps gives SGD -1.0702, Momentum -3.5139, RMSprop -1.0702, Adam -1.0702
SGD, RMSprop and Adam stay in the shallow right-hand well (about -1.07). Momentum builds up speed and rolls over the hill into the deeper left-hand well (about -3.51).
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.
What is a local minimum?
A point lower than all points nearby but not the lowest overall. Gradient descent can stop there because the slope is zero.
How do you escape local minima?
Use momentum, noise from small batches, several random starts or a learning-rate schedule. In very high dimensions many local minima are nearly as good as the best.
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.
- Adam Optimizer VisualizerSee Adam adapt its step per direction on the curved Rosenbrock valley.
- RMSprop Optimizer VisualizerSee RMSprop normalise its steps on a surface with four minima.
- 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.