The paper presents a spiking neural network (SNN) for digit recognition that uses spike-timing-dependent plasticity (STDP), lateral inhibition, and adaptive spiking threshold. Unlike most other SNNs, this network does not rely on a teaching signal or class labels, achieving 95% accuracy on the MNIST benchmark. The network architecture includes leaky integrate-and-fire neurons, conductance-based synapses, and exponential time-dependent weight changes. The authors demonstrate the robustness of the network by varying the learning rule and observing similar performance across different mechanisms. The network's scalability and general applicability are highlighted, making it a promising approach for understanding and simulating complex computations in the mammalian neocortex.The paper presents a spiking neural network (SNN) for digit recognition that uses spike-timing-dependent plasticity (STDP), lateral inhibition, and adaptive spiking threshold. Unlike most other SNNs, this network does not rely on a teaching signal or class labels, achieving 95% accuracy on the MNIST benchmark. The network architecture includes leaky integrate-and-fire neurons, conductance-based synapses, and exponential time-dependent weight changes. The authors demonstrate the robustness of the network by varying the learning rule and observing similar performance across different mechanisms. The network's scalability and general applicability are highlighted, making it a promising approach for understanding and simulating complex computations in the mammalian neocortex.