---
title: Continuation-Based Learning Framework
url: https://www.emergentmind.com/topics/continuation-based-learning-framework
type: topic
---

# Continuation-Based Learning Framework

A continuation-based learning framework is a family of learning formulations in which adaptation proceeds through a structured progression rather than a single static optimization. In current arXiv usage, the phrase appears in several technically distinct settings: continual learning under non-iid task streams, sequential Bayesian and meta-continual learning, curriculum-ordered online class-incremental learning, optimization by continuation in reinforcement learning, and supervised operator learning for analytic continuation in numerical quantum many-body physics [2102.06253] [1902.06494] [2211.15470] [2305.06851] [1905.10430]. Across these uses, the common motif is that learning is mediated by an evolving object—such as a scenario, a posterior, a latent concept distribution, a recurrent state, or a smoothed objective—that carries information from earlier stages into later ones.

## 1. Conceptual range of the term

In the cited literature, “continuation-based” does not denote a single standardized architecture. Instead, it denotes a methodological pattern in which learning is organized along a path: over tasks, over data-distribution shifts, over latent posterior updates, or over progressively less smoothed objectives [2102.06253] [2405.18758] [2310.11952] [2305.06851] [1905.10430].

| Setting | Continued object | Representative formulation |
|---|---|---|
| Continual learning infrastructure | Ordered task stream | Dataset $\rightarrow$ Taskset $\rightarrow$ Scenario |
| Bayesian/meta-continual learning | Posterior or sufficient statistics | Sequential Bayesian update over $\boldsymbol{\omega}$ or $z$ |
| Replay/distillation frameworks | Latent concept distribution or past representation space | GMM-based replay, predictor-based SSL distillation |
| Curriculum and sequence modeling | Class order or recurrent state | Curriculum over classes; CL as forward pass of a sequence model |
| Optimization/analytic continuation | Smoothed objective or inverse operator | Mirror policies; $G(\tau)\mapsto A(\omega)$ |

A useful way to organize the literature is by asking what is being continued. In continual-learning systems, the continued object is usually a task-conditioned data stream or a compact memory of prior tasks. In Bayesian and meta-learning systems, it is a posterior distribution or a fixed-dimensional sufficient statistic. In reinforcement learning, it is a smoothed surrogate of the return. In analytic continuation, it is a learned map from noisy imaginary-time data to spectral density. This suggests that the expression names a shared research style rather than a single problem domain.

## 2. Scenario construction and experimental infrastructure

In continual learning, the central difficulty is that the training distribution is not static, drifts through time, and can induce interference with previously learned knowledge. “Continuum: Simple Management of Complex Continual Learning Scenarios” formalizes this engineering problem and treats data loading itself as a first-class research object [2102.06253]. Its core abstraction separates **Datasets**, **Tasksets**, and **Scenarios**: raw data are wrapped into task-specific subsets, and a scenario is an ordered sequence of tasks constituting the actual continual-learning curriculum.

This separation supports several common regimes. The framework provides **incremental learning**, in which new tasks bring new concepts or classes; **lifelong learning**, in which the class set stays fixed but the input distribution changes; mixed incremental and lifelong settings, also called **NIC**; and transformation-based scenarios such as **Rotation-MNIST** and **Permuted-MNIST**. It also supports metadata-driven datasets such as **CORe50**, where sample identifiers can be used to construct NIC-style streams. The same base dataset can therefore be instantiated as class-incremental, domain-incremental, transformed, or metadata-conditioned streams.

The framework also standardizes evaluation. Its Logger includes **accuracy**, **average incremental accuracy**, **online cumulative performance**, **backward transfer**, **remembering**, **positive backward transfer**, and **model size efficiency**. A central trajectory-sensitive metric is average incremental accuracy,
$$
\frac{1}{T}\sum_{t=1}^{T} A^t,
$$
where $A^t$ is the accuracy after task $t$. The framework further distinguishes **performance metrics**, **behavior metrics**, and **computational metrics**, reflecting the fact that continual learning is assessed over a sequence of models rather than a single terminal checkpoint.

A notable design point is reproducibility. The framework is explicitly motivated by the observation that small errors in preprocessing, task order, or test-set construction can significantly distort continual-learning results. It therefore adopts a deliberately narrow scope—mainly data loading and metrics—described in the paper as a **Unix philosophy**. The implementation is also extendable: each object can be inherited, and new datasets, task partitioning rules, transformations, or supervision signals can be added without rewriting the rest of the pipeline. In this literature, continuation is thus operationalized as controlled scenario construction rather than as a new forgetting-avoidance algorithm.

## 3. Sequential Bayesian, bilevel, and unified optimization formulations

A major theoretical line treats continual learning as sequential Bayesian inference. “A Unifying Bayesian View of Continual Learning” starts from the exact recursive update
$$
p(\boldsymbol{\omega}\mid \mathcal{D}_{1:t}) \propto p(\mathcal{D}_t \mid \boldsymbol{\omega})\, p(\boldsymbol{\omega}\mid \mathcal{D}_{1:t-1}),
$$
and argues that practical continual learning becomes difficult because the posterior is intractable for models such as Bayesian neural networks [1902.06494]. The paper distinguishes **prior-focused** methods, which reuse the previous approximate posterior as the new prior—covering VCL, EWC, SI, Riemannian Walk, and related methods—from **likelihood-focused** methods, which preserve past-task likelihood contributions via replay or generative models, including DGR, pseudo-rehearsal, core-set replay, and the paper’s own **VGR**. It then derives a **hybrid** objective combining both views. An important empirical claim is that prior-focused methods can fail in realistic settings such as **single-headed Split MNIST**, where VCL performs much worse than replay-based approaches and exhibits poorly calibrated uncertainty, whereas VGR behaves more sensibly.

“Bilevel Continual Learning” recasts the same broad problem as a meta-learning-style online optimization with a **dual memory system** [2007.15553]. It uses **episodic memory** $M^{er}$ for replay in the inner problem and **generalization memory** $M^{gm}$ for a validation-like outer problem. For each incoming mini-batch $B_n^t$, the method solves
$$
\min_{\theta}\; L^{outer}\big(\phi^*(\theta), M^{gm}; \theta\big)
\quad \text{s.t.} \quad
\phi^* = \arg\min_{\phi} L^{inner}\big(\phi, B_n^t \cup M_n^{er}; \theta\big).
$$
The fast weights $\phi$ are initialized from the main model $\theta$, adapted on current data plus replay, and then used to update $\theta$ through a first-order approximation. A distillation regularizer reduces bias toward the current task when replay memory is small. The reported benchmarks—**Permuted MNIST**, **Split CIFAR-100**, **Split CUB**, and **Split miniImagenet**—show **BCL-Dual** as generally the strongest variant across **ACC**, **FM**, and **LA**.

“Learning to Continually Learn with the Bayesian Principle” pushes the Bayesian view further by freezing the neural network entirely during continual learning and allowing only a statistical latent-variable model to update [2405.18758]. In **Sequential Bayesian Meta-Continual Learning (SB-MCL)**, continual learning is represented by a posterior over an episode-specific latent variable $z$, while the neural network is meta-trained beforehand to map raw data into the likelihood terms required by that posterior. For a factorized Gaussian posterior, the sequential update is exact:
$$
\Lambda_t = \Lambda_{t-1} + P_t,\qquad
\mu_t = \Lambda_t^{-1}\left(\Lambda_{t-1}\mu_{t-1} + P_t z_t\right).
$$
Because the neural network performs only forward passes during the stream, the framework is explicitly protected from catastrophic forgetting by design. The paper reports substantial efficiency gains in meta-training time, including **classification: OML 6.5 hr, TF 1.2 hr, SB-MCL 40 min**, and analogous advantages in completion, VAE, and DDPM settings.

A separate unification is provided by “A Unified and General Framework for Continual Learning,” which expresses CL as
$$
\mathcal{L}^{CL}
=
\mathcal{L}_{CE}(x,y)
+
\alpha D_{\phi}(h(x), r)
+
\beta D_{\psi}(\theta,\theta_{old}),
$$
with output-space and weight-space Bregman divergences [2403.13249]. Under special choices of $\phi$, $\psi$, and the references $r$ or $\theta_{old}$, the framework recovers EWC, CPR, VCL, NCL, ER, and DER. Its additional **refresh learning** plug-in performs deliberate unlearning followed by relearning of the current mini-batch. Reported gains include **ER on CIFAR-100 Class-IL: $20.98 \to 22.23$** and **DER++ on CIFAR-100 Task-IL: $75.64 \to 77.71$**. These results formalize a recurring tension in the literature: retention is not always identical to preserving every earlier parameter configuration unchanged.

## 4. Representation continuation through replay, distillation, and drift adaptation

Some continuation-based frameworks preserve knowledge primarily in representation space rather than in parameter priors. “Generative Continual Concept Learning” proposes **ECLA**, an autoencoder-classifier architecture in which new forms of a concept are coupled to previously learned forms in a shared latent space [1906.03744]. The encoder $\phi_{\mathbf{v}}$ maps inputs into a discriminative embedding, a decoder $\psi_{\mathbf{u}}$ makes the system generative, and a **Gaussian mixture model (GMM)** over latent clusters approximates a task-independent concept distribution. For later tasks, the method trains on both current labeled samples and pseudo-replayed samples generated from the latent GMM, while aligning current and historical latent distributions using **Sliced Wasserstein Distance (SWD)**. The paper evaluates this on **Permuted MNIST** and sequential **MNIST $\rightarrow$ USPS / USPS $\rightarrow$ MNIST**, with only the first task fully labeled and about **10 labeled examples per class** for the new task in the transfer setting. The reported effects include retained first-task performance, a **jump-start** on the second task, and embeddings that merge same-class instances across domains into shared clusters.

“Self-Supervised Models are Continual Learners” develops **CaSSLe**, which turns a self-supervised loss into a continual distillation mechanism by adding a predictor network $g$ that maps current features into the previous representation space [2112.04215]. Instead of directly constraining $\mathbf{z}$ to equal $\bar{\mathbf{z}}$, CaSSLe learns
$$
g(\mathbf{z}) \approx \bar{\mathbf{z}},
$$
and uses the same SSL objective both for current-task learning and for temporal distillation. The method is applied to **SimCLR**, **MoCoV2+**, **BYOL**, **SwAV**, **Barlow Twins**, and **VICReg**. It uses a **2-layer MLP with 2048 hidden units and ReLU**, sums the SSL and distillation losses without an additional weighting hyperparameter, and adds about **30% extra memory/time** overhead. Reported average linear-evaluation gains are about **6.8%** on class-incremental **CIFAR100**, about **4%** on **ImageNet100**, and about **4.4%** on domain-incremental **DomainNet**.

A different representation-level continuation appears in “Class-Incremental Experience Replay for Continual Learning under Concept Drift,” which explicitly combines continual learning with drift adaptation [2104.11861]. The method maintains a **centroid-driven memory** with replay buffers attached to multiple centroids per class and augments it with a **Reactive Subspace Buffer (RSB)** that tracks local label changes. Replay is purity-aware:
$$
\gamma_C = \tanh\!\left(\beta \frac{c_1-c_2}{c_1+c_2}\right) > r,\qquad \beta=4,
$$
and clusters are split when
$$
\frac{c_1}{c_2}-1.0 < \tau_s.
$$
The paper evaluates **MNIST**, **FASHION**, **SVHN**, **CIFAR10**, and **IMAGENET10** under both stationary and drift settings. In the drift setting, reported normalized average accuracies for **ER-RSB** are **0.9938** on MNIST, **0.9745** on FASHION, **0.9722** on SVHN, **0.9545** on CIFAR10, and **0.9187** on IMAGENET10. A central implication is that a continuation-based learner need not preserve all historical information indiscriminately; it may instead preserve valid subspaces and revise obsolete ones.

## 5. Curriculum ordering and sequence-model formulations

Another research line treats continuation as the temporal ordering of classes or tasks. “Learning to Learn: How to Continuously Teach Humans and Machines” studies **online class-incremental continual learning** in which each example is shown once and the curriculum is the order of classes, $T=(t_1,\dots,t_N)$ [2211.15470]. Curriculum quality is evaluated by **$\alpha$**, the average accuracy over all seen classes, and **$\beta$**, the drop in first-task accuracy between the first and last task, combined into
$$
\mathcal{F}=\frac{2}{\beta + \frac{1}{\alpha}}.
$$
The paper proposes a feature-based **Curriculum Designer (CD)** using SqueezeNet features pretrained on ImageNet, with **500 images per class** to compute prototypes. The ranking heuristic favors a centrally located first class, larger inter-class distances in the early middle portion, and late classes similar to the first class as a replay-like reinforcement mechanism. Human experiments on the **Novel Object Dataset (NOD)** retained **169** MTurk subjects after filtering, for **34,848 test trials**, and also included an in-lab study with **60 subjects**. The supplement reports **Spearman correlations on MNIST paradigm-I** of **0.26** between algorithms, **0.08** between algorithm and CD, and **0.0002** between algorithm and random ranking; curriculum discrepancy $\mathcal{H}$ decreases from random-human to algorithm-human by about **0.13**. In this literature, continuation is the deliberate progression of concepts through a teaching order.

“Recasting Continual Learning as Sequence Modeling” goes further by replacing inner-loop parameter updates with state updates inside a sequence model [2310.11952]. A continual learner is written as a functional
$$
H_\eta : (\mathcal X \times \mathcal Y) \times (\mathcal X \to \mathcal Y) \to (\mathcal X \to \mathcal Y),
$$
and the stream-processing version is
$$
f_{\theta_t} = G_\eta(((x_1,y_1),\dots,(x_t,y_t))).
$$
In the proposed formulation, the continual-learning process becomes the **forward pass of a sequence model**, with the hidden state $h_t$ carrying accumulated information from the stream. Decoder-only **Transformers** and kernel-based efficient Transformers such as **Linear Transformer** and **Performer** are then used as meta-continual learners. For kernel-based attention, each head maintains recurrent statistics
$$
S_t = S_{t-1} + \phi(k_t)v_t^\top,\qquad
z_t = z_{t-1} + \phi(k_t),
$$
yielding linear-time state updates. The paper reports experiments on **seven benchmarks**, covering both classification and regression, with default **$K=20$** tasks per episode and also **$K=100$** for longer episodes. This suggests a shift from parameter continuation to state continuation: the learner is not repeatedly re-optimized but instead accumulates episode-specific state through causal sequence processing.

## 6. Optimization by continuation and analytic continuation beyond standard continual learning

Outside continual-learning benchmarks, continuation-based learning also denotes objective smoothing and learned inverse operators. “Policy Gradient Algorithms Implicitly Optimize by Continuation” formulates direct policy optimization in the **optimization by continuation** framework [2305.06851]. Instead of optimizing a nonconvex objective directly, one optimizes a family of smoother surrogate objectives
$$
f^p(x)=\mathbb{E}_{y\sim p(\cdot\mid x)}[g(y)].
$$
For reinforcement learning, the continuation of the policy return is defined by perturbing policy parameters over time, $\theta_t \sim q(\cdot \mid \theta,\Lambda(h_t))$, and the resulting smoothed objective is shown to equal the return of a **mirror policy**
$$
\eta'_\theta(a\mid h)=
\mathbb{E}_{\theta' \sim q(\cdot\mid \theta,\Lambda(h))}
\left[\eta_{\theta'}(a\mid h)\right].
$$
For affine deterministic policies under Gaussian perturbations, the mirror policy becomes Gaussian with covariance
$$
\Sigma'_\theta(s)=\nabla_\theta \mu_\theta(s)^\top \Lambda(s)\nabla_\theta \mu_\theta(s).
$$
The paper interprets stochastic exploration and entropy regularization as mechanisms that maintain this smoothing, and argues that policy variance should be a history-dependent continuation parameter chosen to avoid poor local extrema rather than merely to maximize immediate return.

“Analytic Continuation of Noisy Data Using Adams Bashforth ResNet” uses the term in yet another sense: learning the inverse operator that maps noisy imaginary-time data to a spectral function in quantum many-body physics [1905.10430]. The forward relation is the Fredholm integral equation
$$
G(\tau)=\int_{-\infty}^{\infty}
\frac{e^{-\tau\omega}}{1+e^{-\beta\omega}}
A(\omega)\,d\omega,
$$
and the baseline **MaxEnt** method minimizes
$$
Q=\frac{1}{2}\chi^2[A]-\alpha S[A].
$$
The proposed **AB-ResNet** replaces iterative inverse-problem solving with a supervised map $G(\tau)\mapsto A(\omega)$ learned by a residual network interpreted as an ODE discretization. Standard ResNet corresponds to **AB1**, and the paper proposes higher-order **AB2** and **AB3** multistep residual updates. Synthetic training data are generated from Gaussian-mixture spectra, noise levels are **$10^{-5}$, $10^{-3}$, $10^{-2}$**, the imaginary-time signal is represented by **64 Legendre polynomial coefficients**, and the reported dataset sizes are **100,000** for training, **1000** for validation, and **1000** for testing. Mean absolute errors are **$6.8\times 10^{-4}$** for AB1, **$3.8\times 10^{-4}$** for AB2, and **$2.6\times 10^{-4}$** for AB3. Reported CPU times are on the order of **$10$ seconds** for AB-ResNet versus **$10^3$ seconds** for MaxEnt. Here, continuation does not mean task sequencing at all; it means learning a stable approximation to an ill-posed continuation operator.

Across these literatures, the term “continuation-based learning framework” refers to a common structural principle rather than a single canonical method. What is continued may be a task stream, a posterior, a latent concept model, a curriculum, a recurrent state, a smoothed control objective, or an inverse physical operator. A recurring point of disagreement concerns whether historical information should always be preserved: prior-focused Bayesian CL, replay systems, and infrastructure frameworks emphasize retention, whereas concept-drift adaptation and refresh learning explicitly allow outdated information to be discarded or unlearned [1902.06494] [2104.11861] [2403.13249]. The literature therefore presents continuation not as a uniform algorithmic recipe, but as a broad technical paradigm for organizing learning over structured progression.

Source: https://www.emergentmind.com/topics/continuation-based-learning-framework