Synaptic is an architecture-free neural network library for JavaScript that supports various neural network types including multilayer perceptrons, LSTM networks, liquid state machines, and Hopfield networks. It includes a built-in trainer capable of training any given network architecture, with built-in tasks like solving XOR, completing Distracted Sequence Recall, and passing the Embedded Reber Grammar test. The library implements generalized algorithms derived from Derek D. Monner’s paper on LSTM-like training for second-order recurrent networks.