Associative Data Storage and Retrieval in Neural Networks

By Günther Palm, Friedrich T. Sommer. 1996

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Edgar Körner
2. Ursula Körner
3. Andreas Knoblauch
4. Friedrich T. Sommer
5. German Ignacio Parisi
6. Günther Palm

Citations in the corpus

• Structural Synaptic Plasticity Has High Memory Capacity and Can Explain Graded Amnesia, Catastrophic Forgetting, and the Spacing Effect (2014, background): “By comparison, even with real-valued synapses that have an infinite number of states, Hebbian-type weight plasticity without structural plasticity achieves less than one bit per synapse [72,73,139,140].”
• Memory Capacities for Synaptic and Structural Plasticity (2010, methodology): “Alternative approaches based on signal-to-noise theory (e.g., Dayan and Willshaw, 1991; Palm and Sommer, 1996) are better suited for finite feed-forward networks with asymmetric weight matrix but require Gaussian assumptions on the distribution of dendritic potentials which may lead to inaccurate…”
• Zip nets: Efficient associative computation with binary synapses (2010, background): “1 (left panel), learning of synaptic potentials is determined by four learning increments r1 := R(0, 0), r2 := R(1, 0), r3 := R(0, 1), and r4 := R(1, 1).”
• Improved bidirectional retrieval of sparse patterns stored by Hebbian learning (1999, unknown): “5 (Palm and Sommer, 1995), and asymptotically, the Willshaw model is the most efficient NAM, see the model comparison in Sommer and Dayan (1998).”
• Optimal synaptic learning in non-linear associative memory (2010, methodology): “In summary, forMpq → ∞ the SNR r ≈ m/(Mq(1−q)) of optimal Bayesian learning is identical to the asymptotic SNR of linear learning with the optimal covariance rule (e.g., see ρCovariance3 in [33] p259; or eq.”
• Neural Associative Memory with Optimal Bayesian Learning (2011, background): “…does the performance drop from 0.72 bps to the capacity of the original (nonsparse) Hopfield network (e.g., C = 0.14 bps in Hopfield, 1982; Hertz, Krogh, & Palmer, 1991; Palm & Sommer, 1996, or, as we will see below, C = 0.33 bps for the heteroassociative feedforward networks considered here).”