An optician turns the frame, adjusts the focus, and a blurry spot becomes a sharp image. The new algorithm GOAL works much the same way: it rotates the data cloud, compresses redundant dimensions, and brings into focus what truly matters.
When there are more features than observations, standard methods are like trying to make out a face through a fogged-up window: noise drowns out the signal. GOAL takes a different approach. It tackles three tasks at once: it rotates the coordinate system to discard empty directions, partitions the space into cells, and fits a classifier so that predictions match the true labels. Each step is computed analytically, without lengthy tuning, and convergence is proven mathematically.
The secret is that the algorithm does not try to grasp the ungraspable. The rotation is computed via singular value decomposition, the same technique used to compress photographs. Cell membership is assigned by nearest center with a label correction, and label probabilities are updated by frequencies. All this comes together in a single loop, where each iteration sharpens the picture a bit more.
In reconstructing gene activity networks from 182 genes, GOAL was the best in 82.7% of 910 scenarios, taking 0.019 seconds where random forest took 9.84 seconds and LSTM took 39.88. This speed advantage means analyses can run on a standard laptop rather than a cluster.
The optician puts the lens back in place: now you can see not just the outlines but also the details. GOAL does the same with data that are few and noisy. And where such tasks once required expensive computation, now focus is achieved in hundredths of a second.
For the El Niño forecast, 295,200 features were used with only 720 observations, which is like trying to guess a melody from 720 notes when the score has 295,200 lines.