Digital Futures

Experience unfolds continuously, with current states constrained by prior activity and changing context. Domenico Guarino tests whether this continuity appears as reproducible neural trajectories rather than isolated events. Neuronal population events form persistent trajectories whose transitions depend on prior states and other brain areas. In human EEG, microstate sequences exhibit dependencies beyond dwell time and first-order Markov predictions, and stimuli systematically reshape those dependencies. Linking ensemble and EEG dynamics may reveal classes of individual responses and inform general AI models predicting responses to stimuli.

Pierre Baldi develops a general theory in which a neuron is balanced when its input and output weights have equal regularization cost. The theory extends beyond ReLU networks with L2 penalties to other activations, Lp regularizers, recurrent and convolutional architectures, and mixed networks. Ordinary gradient descent need not preserve balance; regularized learning should converge toward it. Local scaling and balancing operations lead to a stochastic algorithm whose unique global solution follows from strictly convex optimization on an architecture-dependent linear manifold. Simulations suggest that balancing before or during training improves speed and performance. Connections to overparameterization, group symmetries and neurobiological or neuromorphic networks are discussed.