The fused Lasso adds an L1 norm penalty on adjacent coefficient differences to a squared-error objective, optionally together with an L1 norm penalty on the coefficients themselves. The differences penalty encourages neighbouring coefficients to coincide.
New to topics? Read the docs here!