Isomorphic Functionalities between Ant Colony and Ensemble Learning: Part III -- Gradient Descent, Neural Plasticity, and the Emergence of Deep Intelligence
Authors: Ernest Fokoué, Gregory Babbitt, Yuval Levental
Organizations: School of Mathematics and Statistics, Rochester Institute of Technology · Gosnell School of Life Sciences, Rochester Institute of Technology · Center for Imaging Science, Rochester Institute of Technology
In Parts I and II of this series, we established isomorphisms between ant colony decision-making and two major families of ensemble learning: random forests (parallel, variance reduction) and boosting (sequential, bias reduction). Here we complete the trilogy by demonstrating that the fundamental learning algorithm underlying deep neural networks -- stochastic gradient descent -- is mathematically isomorphic to the generational learning dynamics of ant colonies. We prove that pheromone evolution across generations follows the same update equations as weight evolution during gradient descent, with evaporation rates corresponding to learning rates, colony fitness corresponding to negative loss, and recruitment waves corresponding to backpropagation passes. We further show that neural plasticity mechanisms -- long-term potentiation, long-term depression, synaptic pruning, and neurogenesis -- have direct analogs in colony-level adaptation: trail reinforcement, evaporation, abandonment, and new trail formation. Comprehensive simulations confirm that ant colonies trained on environmental tasks exhibit learning curves indistinguishable from neural networks trained on analogous problems. This final isomorphism reveals that all three major paradigms of machine learning -- parallel ensembles, sequential ensembles, and gradient-based deep learning -- have direct analogs in the collective intelligence of social insects, suggesting a unified theory of learning that transcends substrate. The ant colony, we conclude, is not merely analogous to learning algorithms; it is a living embodiment of the fundamental principles of learning itself.
Figures & tables
Neural Network
Ant Colony
Network weights w
Pheromone configuration τ
Training epoch t
Generation g
Mini-batch Bt
Recruitment wave within generation
Forward pass
Ant foraging guided by pheromone
Loss function L(w)
Negative colony fitness −F(τ)
Gradient ∇L(wt)
Fitness gradient ∇τF(τg)
Table 1: Correspondence between neural network training and generational ant colony learning
Figure 1: The gradient descent isomorphism. After normalization to [0,1] , the ant colony error signal (defined as 1−Fnorm ) and the neural network loss follow nearly identical trajectories over 50 generations/epochs, illustrating the correspondence ρGACL≅ηNN .
Figure 2: Isomorphic evolution of pheromone concentrations and neural network weights. (a) Pheromone levels for five foraging sites over 50 generations; the best site (highest final concentration) emerges through competitive reinforcement. (b) Five representative weight magnitudes over 50 training epochs; a dominant weight emerges through gradient-driven competition. Both systems show identical dynamics of initial exploration followed by convergence to a winner.
Figure 3: Gradient dynamics in both systems. (a) Ant colony: the error signal (negative fitness) and the magnitude of the inter-generational fitness change ∣ΔF∣ both decrease as the colony converges. (b) Neural network: the loss signal and gradient norm show an analogous pattern. In both cases the gradient magnitude is large when the error is large and decays as the system approaches its optimum, confirming the isomorphism at the level of gradient dynamics.
Dataset
Neural Network
GACL
Colony-Net
Iris (easy)
1.000±0.000
1.000±0.000
1.000±0.000
Iris (hard)
0.922±0.055
0.872±0.060
0.897±0.047
mtcars
0.742±0.198
0.817±0.131
0.779±0.139
Swiss
0.728±0.167
0.817±0.178
0.772±0.121
USArrests
0.830±0.113
0.910±0.079
0.870±0.080
Average
0.844
0.883
0.864
Table 2: Classification accuracy (mean ± SD over 20 replicates) on built-in R datasets. GACL uses centroid-based site qualities and colony decisions; Colony-Net averages the predictions of both systems.
Figure 4: Uniform convergence of GACL. With observation noise set to zero, the trajectory variance decreases as a power law in colony size N . The fitted exponent of −1.44 ( R2=0.95 ) is consistent with the O(1/N) rate predicted by the law of large numbers applied to the multinomial ant allocation, confirming that the GACL learning dynamics converge to a deterministic limit.
Figure 5: Visual illustration of uniform convergence. Each panel shows 20 independent GACL replicates (faint lines) and their mean (bold line) for increasing colony sizes. At N=10 trajectories are highly variable; by N=1000 they are tightly clustered, illustrating the convergence to a deterministic limit.
Figure 6: Learning curves for the ant colony (GACL) and neural network across 20 independent replicates. After normalization to [0,1] , the mean trajectories show similar convergence dynamics. Faint lines show individual replicates; shaded bands indicate ±1 standard error.
Figure 7: Learning rate sensitivity. Final normalized performance (averaged over 15 replicates) as a function of the learning rate η (neural network) and evaporation rate ρ (ant colony). Both systems exhibit a similar inverted-U profile, peaking at moderate rates and degrading at extreme values. Error bars: ±1 SE.
Figure 8: Convergence rates across problem complexity. (a) Linear decision boundary / well-separated sites: both systems converge rapidly. (b) Quadratic boundary / moderate separation: convergence is slower but trajectories remain indistinguishable. (c) Complex non-linear boundary / subtle site differences: both systems converge slowly with similar variance. Shaded bands: ±1 SE over 15 replicates.
Figure 9: Plasticity and adaptation to environmental change. At generation/epoch 25 the optimal site (for the ant colony) and decision boundary (for the neural network) shift abruptly. Both systems show an immediate performance drop followed by recovery, with statistically indistinguishable adaptation rates. Shaded bands: ±1 SE over 15 replicates.
Figure 10: Noise robustness. Normalized final performance (averaged over 15 replicates) as a function of increasing noise level σ . For the ant colony, noise is applied to site quality observations; for the neural network, noise is applied as label flipping. Both systems degrade gracefully and at comparable rates, consistent with the exponential decay model of Eq. ( 26 ). Error bars: ±1 SE.
In Part I of this series, we established a rigorous mathematical isomorphism between ant colony decision-making and random forest learning, demonstrating that variance reduction through decorrelation is a universal principle shared by biological and computational ensembles. Here we turn to the complementary mechanism: bias reduction through adaptive weighting. Just as boosting algorithms sequentially focus on difficult instances, ant colonies dynamically amplify successful foraging paths through pheromone-mediated recruitment. We prove that these processes are mathematically isomorphic, establishing that the fundamental theorem of weak learnability has a direct analog in colony decision-making. We develop a formal mapping between AdaBoost's adaptive reweighting and ant recruitment dynamics, show that the margin theory of boosting corresponds to the stability of quorum decisions, and demonstrate through comprehensive simulation that ant colonies implementing adaptive recruitment achieve the same bias-reduction benefits as boosting algorithms. This completes a unified theory of ensemble intelligence, revealing that both variance reduction (Part I) and bias reduction (Part II) are manifestations of the same underlying mathematical principles governing collective intelligence in biological and computational systems.
Ernest Fokoué, Gregory Babbitt, Yuval Levental
School of Mathematics and Statistics, Rochester Institute of Technology · Gosnell School of Life Sciences, Rochester Institute of Technology · Center for Imaging Science, Rochester Institute of Technology
Social insect colonies and ensemble machine learning methods represent two of the most successful examples of decentralized information processing in nature and computation respectively. Here we develop a rigorous mathematical framework demonstrating that ant colony decision-making and random forest learning are isomorphic under a common formalism of stochastic ensemble intelligence. We show that the mechanisms by which genetically identical ants achieve functional differentiation -- through stochastic response to local cues and positive feedback -- map precisely onto the bootstrap aggregation and random feature subsampling that decorrelate decision trees. Using tools from Bayesian inference, multi-armed bandit theory, and statistical learning theory, we prove that both systems implement identical variance reduction strategies through decorrelation of identical units. We derive explicit mappings between ant recruitment rates and tree weightings, pheromone trail reinforcement and out-of-bag error estimation, and quorum sensing and prediction averaging. This isomorphism suggests that collective intelligence, whether biological or artificial, emerges from a universal principle: \textbf{randomized identical agents + diversity-enforcing mechanisms → emergent optimality}.
Ernest Fokoué, Gregory Babbitt, Yuval Levental
School of Mathematics and Statistics, Rochester Institute of Technology · Gosnell School of Life Sciences, Rochester Institute of Technology · Center for Imaging Science, Rochester Institute of Technology
Understanding how stochastic gradient descent learns internal representations in wide neural networks trained on rich target functions is a central problem in machine learning. We study this question in two-layer neural networks learning target functions represented by networks of the same architecture, with arbitrary nonlinearity, and with network width and input dimension growing proportionally. In this regime, standard approaches for describing the learning dynamics require tracking a diverging number of overlaps, which measure correlations between learned and target features, leaving no obvious description in terms of a few order parameters. Here, we develop a mesoscopic description of learning by characterizing the distribution of overlaps. In the early stages of learning, we find that a Gaussian description of overlap statistics reduces the learning dynamics to a few collective variables that track learning of the target function's linear and quadratic components. We then estimate the timescale of specialization, when Gaussian closure breaks down and individual learned features begin to strongly correlate with specific target features. We find that specialization in extensive-rank networks is accelerated via a quadratic shielding mechanism, whereby learning the linear and quadratic components of the target function filters gradient fluctuations. We finally describe how the activation function determines the pattern of specialization and characterize the subsequent refinement of learned features. Together, these results show how collective learning gives way to individual feature discovery and clarify when learning in large neural networks admits a simple statistical description.
Shu Zhou, Jean Barbier, Sebastian Goldt
International School for Advanced Studies (SISSA) Via Bonomea 265, 34136 Trieste, Italy · The Abdus Salam International Centre for Theoretical Physics (ICTP) Strada Costiera 11, 34151 Trieste, Italy