Auxiliary Learning for Deep Multi-task Learning

Bohan Zhuang; Chunhua Shen; Hao Chen; Wei Yin; Yifan Liu

REVIEW 3 major objections 5 minor 47 references

Auxiliary Learning for Deep Multi-task Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Auxiliary modules supervised by task-specific losses improve hard-parameter-sharing multi-task networks with no inference cost.

desk verdict Useful training-only auxiliary modules for MTL, but the architecture-specific claim is undercut by a confounded deep-supervision control; still worth peer review and a matched comparison. read the letter →

arxiv 1909.02214 v2 pith:XKKAODB4 submitted 2019-09-05 cs.CV cs.LG

Yifan Liu , Bohan Zhuang , Chunhua Shen , Hao Chen , Wei Yin This is my paper · ORCID

classification cs.CVcs.LG

keywords multi-tasklearninghardparametersharingauxiliarymoduleinductivebiassemanticsegmentationdepthestimationsurfacenormalpredictionneuralarchitecturesearch

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Multi-task networks that share one encoder across tasks are cheap to deploy but hard to train, because the different task losses pull the shared weights in competing directions. This paper claims that attaching temporary auxiliary modules to the shared hidden layers during training—each module reading multi-level features and supervised by one task's loss—regularizes the shared encoder and improves every task, at no extra inference cost because the modules are removed after training. On NYUD-v2, the method lifts joint semantic segmentation and depth estimation from 42.9 to 47.9 mIoU and from 0.143 to 0.127 relative error, and it also improves a three-task version on SUNRGBD. A careful reader would care because this is a training-only regularizer that makes hard parameter sharing competitive with single-task models while keeping deployment simple.

What carries the argument

The key machinery is the auxiliary module, a small side network formed from adaptors and aggregators. In the basic form, each adaptor is a $1\times1$ convolution followed by batch normalization and ReLU, and the aggregator is either per-pixel summation or channel-wise concatenation; the module consumes $P$ intermediate feature maps from the shared encoder and outputs a prediction under the auxiliary task loss. Its role is to give the shared parameters extra gradient flow during backpropagation, acting as a regularizer that introduces inductive bias from the auxiliary task. For the searched variant, an LSTM controller with reinforcement learning chooses the adaptor operations, aggregation operations, and input locations for each task, using PPO to maximize a validation reward.

What would settle it

Train the same multi-task network with the auxiliary module replaced by direct deep supervision at the same intermediate feature maps, using the same auxiliary losses, loss scales, and training schedule. If this deep-supervision baseline matches the auxiliary-module results, the claimed benefit of the multi-level auxiliary architecture is not supported.

Watch

Extended reading notes

Core claim

The central claim is that the optimization difficulty of hard parameter sharing in multi-task learning can be relieved by explicit auxiliary modules that mimic soft parameter sharing during training. Each auxiliary module takes outputs from several shared hidden layers, passes them through trainable adaptors, aggregates the adapted features, and is supervised by the corresponding auxiliary task loss. Because the auxiliary module's parameters attach to the shared layers, its loss contributes extra gradient terms to the shared parameters, introducing a hierarchical inductive bias that helps balance shared and task-specific representations. After training, the auxiliary modules are discarded, so inference uses only the original multi-task network. The paper reports consistent gains across backbones, decoder designs, and two- and three-task settings, including results better than single-task baselines.

Load-bearing premise

The load-bearing premise is that the extra gradients from the auxiliary modules, which are computed from the same labels as the main tasks, push the shared encoder toward better representations rather than merely reweighting or duplicating the original supervision.

Editorial extensions

If this is right

Hard-parameter-sharing multi-task models can match or beat single-task baselines without any change to the inference-time architecture, so deployment speed and parameter count remain unchanged.
An auxiliary module for one task can improve that task even when the network is initialized from another task's pretrained weights, and auxiliary modules for related tasks such as surface normal and depth reinforce each other.
The gains appear across decoder variants and shared backbones, including MobileNetV2 and ResNet-50, so the training-time regularizer transfers across main architectures.
Automatically searching the auxiliary module architecture yields larger gains than the hand-designed basic module, and the search is cheap because it is restricted to the auxiliary module rather than the whole network.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Because the auxiliary loss uses exactly the same ground truth as the main task, the reported gains cannot come from extra information; the likely active ingredient is the architectural pathway and multi-level aggregation. An ablation comparing the module with plain deep supervision at the same layers would isolate that ingredient.
The mechanism is not specific to dense prediction: any shared-encoder multi-task model with conflicting gradients should benefit from a training-only auxiliary pathway, so the method is a plausible plug-in for audio or NLP multi-task systems, though the paper does not test those settings.
If the architecture of the auxiliary module is what matters, one could search a single universal auxiliary cell on a small proxy pair of tasks and transplant it to new task combinations, avoiding a per-dataset search; this is a testable extension the paper leaves open.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

The paper proposes a training-only auxiliary module for hard-parameter-sharing MTL, with the module removed at inference, and shows consistent gains across backbones and datasets. The specific combination — multi-level auxiliary inputs, trainable adaptors/aggregators, and NAS over the module — is new relative to the deep supervision literature.

The empirical work is substantial. The paper covers MobileNetV2 and ResNet-50, two and three tasks, NYUD-v2 and SUNRGBD, and reports improvements over Joint baselines that often exceed single-task performance. The objective in Eq. (3) is simple and the derivation is sound. That said, the central claim — that the multi-level architecture, not merely extra same-task supervision, explains the gains — is undercut by a confounded control. The deep supervision baseline in Sec. 4.2 uses a loss scale of 0.1, while the auxiliary losses in Eq. (3) are unscaled (coefficient 1). So the comparison conflates architecture with gradient magnitude. The paper even argues from the limited gain of the 0.1-scaled deep supervision that its architecture matters, but no experiment varies the auxiliary loss weight or tests a simple per-layer classifier at weight 1.0. That is a real gap.

There are also the usual empirical soft spots: no error bars or statistical tests, no code release, and the NAS search is only described at a high level. Those are not fatal but they do limit how much confidence one can place in the individual numbers.

If I were asked to referee this, I'd send it out. The idea is useful and testable, and the confound is fixable with a matched deep supervision baseline. A good referee could ask for that addition plus error bars, and the paper would come back much stronger.

Referee Report

3 major / 5 minor

Summary. The paper proposes an auxiliary learning strategy for hard-parameter-sharing multi-task networks. During training, auxiliary modules are attached to intermediate outputs of the shared layers and are supervised by the same task losses as the main tasks, with the joint objective given in Eq. (3). At test time the auxiliary modules are removed, so the deployed network is the original hard-sharing model with no added inference cost. The method is evaluated on NYUD-v2 and SUNRGBD for semantic segmentation, depth estimation, and surface normal prediction, using MobileNetV2 and ResNet-50 backbones. The paper reports consistent improvements over joint-training baselines, comparisons with loss-weighting and prior MTL methods, and an automatically searched auxiliary module via reinforcement learning.

Significance. If the reported results hold, the method is a simple and practical training-only regularizer that improves hard-parameter-sharing MTL without changing the inference architecture. The paper's strengths include a clean objective with no tuned task weights, broad empirical coverage across backbones, datasets, and numbers of tasks, an efficient NAS procedure for the auxiliary module, and direct comparisons with Kendall et al.'s uncertainty weighting and other MTL baselines. The main weakness is that the central claim about the auxiliary module's architecture is not cleanly isolated from the scale of the auxiliary losses, so the mechanism attributed to the design is not yet fully supported.

major comments (3)

[Sec. 4.2, Eq. (3), Fig. 3] The comparison between Deep supervision and Auxi is confounded by the scale of the auxiliary losses. Deep supervision is defined as adding losses "at the end of each downsampling layer, and sum them all with scale of 0.1" (Sec. 4.2), whereas Eq. (3) adds the auxiliary losses with no coefficient, i.e., effective weight 1.0. Therefore the observation that Deep supervision contributes little while Auxi improves both tasks cannot be attributed to the multi-level auxiliary module architecture; it may simply reflect a larger auxiliary gradient magnitude. To support the paper's mechanism claim, please report a matched ablation, for example deep supervision with per-loss weight 1.0 and/or Auxi with auxiliary losses scaled by 0.1, using the same set of tapped layers.
[Tables 1-4] The empirical conclusions rest on single-run point estimates with no error bars or significance tests. Several margins are small, e.g., Table 2 reports 35.02 vs 35.71 mIoU for Auxi-both with conv1x1 vs conv3x3, and Table 4 reports 53.7 vs 54.0 mIoU for Joint vs Auxi-T2 on segmentation. Without repeated runs or a significance test, these differences could be within run-to-run noise. Please report means and standard deviations over at least three runs, or an equivalent significance test, for the main comparisons in Tables 1-4 and Fig. 3.
[Sec. 3.2, Sec. 4.2] The NAS component is evaluated without specifying the meta-train/meta-val split, the number of architectures sampled per search, or the variance of the search. Since the reported 1.9% mIoU gain of Auxi-NAS over Auxi-both (Fig. 3) is a selection result based on the meta-val reward, it is unclear whether this gain is a genuine property of the searched architecture or an artifact of selection bias. Please provide the split details and, ideally, evaluate the top-k searched structures or repeat the search to demonstrate stability.

minor comments (5)

[Sec. 4.2] The phrase "we get a significant 1.9% improvement" uses "significant" in the non-statistical sense; please replace it with "numerical improvement" or provide a significance test to justify the statistical meaning.
[Eq. (3)] The notation for auxiliary-module parameters is inconsistent: the text uses θAt, while Eq. (3) writes θA1, ..., θAT. Please unify the notation to make clear which parameters belong to the t-th auxiliary module.
[References] References He et al. 2016a and He et al. 2016b are the same paper (Deep Residual Learning for Image Recognition) and should be merged into a single citation.
[Fig. 3] Figure 3 is difficult to read because the markers and labels are small; a table with exact numerical values alongside the figure would make the comparisons clearer.
[Sec. 4.6] The sentence "we observe the auxiliary learning apparently boosts the performance" is vague; please state the quantitative improvements and note whether they are consistent across the metrics reported in Table 4.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the auxiliary losses are extra training objectives, not fitted predictions; the NAS search is validation-based model selection.

full rationale

The paper's derivation chain is Eq. (3): the total objective is a sum of the per-task losses on the main heads plus the same per-task losses evaluated on auxiliary module outputs. Nothing in that objective is fitted to the reported test metrics. The reported gains (e.g., ResNet-50-Auxi-NAS 47.9 mIoU vs ResNet-50-Joint 42.9 mIoU, Table 3) are measured on the NYUD-v2 test set after training; the auxiliary module is removed at inference, so the result is an empirical comparison, not a quantity that equals its input by construction. The NAS procedure selects among auxiliary architectures by validation reward and then reports test performance; this is model selection, not circularity. The only self-citations (Nekrasov et al. 2019 for the search space and acceleration) are used as engineering tools and are not load-bearing for the central claim, since the basic auxiliary module without NAS already improves over Joint (Fig. 3, AT1/AT2/ATboth). The skeptic's concern about deep supervision is a valid experimental-control issue: the deep-supervision baseline uses losses 'sum them all with scale of 0.1' (Sec. 4.2) while Eq. (3) applies auxiliary losses with no coefficient, so the auxiliary-module benefit is not cleanly isolated from loss-scale. However, an unmatched baseline is a confound, not a circular reduction; no equation or fitted parameter is equivalent to the claimed result. Accordingly, the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. The auxiliary module is a network component. The main free choices are the searched architecture and the fixed loss weights; the key assumptions are the relatedness of tasks and the reliability of the validation-based NAS selection.

free parameters (3)

Auxiliary module architecture = Searched structure shown in Figure 4
The architecture is selected by reinforcement learning to maximize validation reward, which is a model selection process that the reported test results depend on.
Number of auxiliary cells P = Set by the backbone (e.g., 5 for MobileNetV2, 4 for ResNet-50)
P equals the number of downsampling layers or residual blocks, and is not systematically varied.
Loss weighting coefficients alpha_t = 1 for all tasks
The paper explicitly sets all task weights to 1, which is a fixed choice that may affect the reported improvements.

assumptions (4)

standard math The chain rule for computing gradients of the objective in Eq. (3) is assumed.
Used to argue that auxiliary losses add gradient terms to the shared parameters.
domain assumption The auxiliary tasks share useful representation with the main tasks.
The method relies on relatedness between segmentation, depth, and normal estimation; it is not tested on unrelated tasks.
domain assumption The pre-trained segmentation network (Nekrasov et al. 2018b) produces coarse labels of sufficient quality for the NYUD-v2-expansion dataset.
The expansion dataset is used to pre-train models in most experiments, and baselines are also pre-trained on it, so the effect is shared but the data quality is assumed.
ad hoc to paper The RL controller's reward (geometric mean of evaluation metrics) on the meta-val set is a reliable proxy for test performance.
This is a standard NAS assumption but is not proven; it is load-bearing for the claim that the searched architecture is better than the basic modules.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Auxiliary Learning for Deep Multi-task Learning." pith.science (2026). https://pith.science/paper/XKKAODB4

@misc{pith2026190902214,
  author       = {Pith},
  title        = {Pith review of: Auxiliary Learning for Deep Multi-task Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XKKAODB4}},
  note         = {Machine review of arXiv:1909.02214}
}

read the original abstract

Multi-task learning (MTL) is an efficient solution to solve multiple tasks simultaneously in order to get better speed and performance than handling each single-task in turn. The most current methods can be categorized as either: (i) hard parameter sharing where a subset of the parameters is shared among tasks while other parameters are task-specific; or (ii) soft parameter sharing where all parameters are task-specific but they are jointly regularized. Both methods suffer from limitations: the shared hidden layers of the former are difficult to optimize due to the competing objectives while the complexity of the latter grows linearly with the increasing number of tasks. To mitigate those drawbacks, this paper proposes an alternative, where we explicitly construct an auxiliary module to mimic the soft parameter sharing for assisting the optimization of the hard parameter sharing layers in the training phase. In particular, the auxiliary module takes the outputs of the shared hidden layers as inputs and is supervised by the auxiliary task loss. During training, the auxiliary module is jointly optimized with the MTL network, serving as a regularization by introducing an inductive bias to the shared layers. In the testing phase, only the original MTL network is kept. Thus our method avoids the limitation of both categories. We evaluate the proposed auxiliary module on pixel-wise prediction tasks, including semantic segmentation, depth estimation, and surface normal prediction with different network structures. The extensive experiments over various settings verify the effectiveness of our methods.

Figures

Figures reproduced from arXiv: 1909.02214 by the authors.

**Figure 2.** (a) Search space of NAS. (b) Controller for generating the single auxiliary cell for the [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Performance of different training strategies. We report the depth prediction and semantic segmentation results on the NYUD-v2. Top-right is better. We can see that adding auxiliary network can significantly boost the performance, even better than that of the single task. 4.2 EFFECT OF DIFFERENT TRAINING STRATEGIES In this section, we explore the effect of different training strategies on multi-task learning with se… view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: The auxiliary modules sampled by the reinforcement learning. We show the detailed struc [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]

**Figure 5.** Figure 5: Training curves. Base: jointly train two tasks. Auxi: adding a single auxiliary module supervised by the depth estimation. Here we show three samples of the average gradients for different layers (a-c) and the training loss curves for the depth estimation task. We can…

**Figure 6.** Figure 6: Visualization results on NYUD-v2. (a) and (d): Semantic segmentation and depth estimation results of ResNet50-joint. (b) and (e): Semantic segmentation and depth estimation results of ResNet50-Auxi-all. (c) and (f): Ground truth. With the help of auxiliary learning, t…

**Figure 7.** Figure 7: Visualization results on SUNRGBD. (a): Input image. (b): Predicted surface normal. (c): Ground truth surface normal. (d): Predicted semantic segmentation results. (e): Ground truth semantic segmentation results. (f): Predicted depth results. (g): Ground truth depth res…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 40 canonical work pages

[1]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...
[2]

A bayesian/information theoretic model of learning to learn via multiple task sampling

Jonathan Baxter. A bayesian/information theoretic model of learning to learn via multiple task sampling. Machine learning, 28 0 (1): 0 7--39, 1997

work page 1997
[3]

Efficient architecture search by network transformation

Han Cai, Tianyao Chen, Weinan Zhang, Yong Yu, and Jun Wang. Efficient architecture search by network transformation. In Proc. AAAI Conf. on Arti. Intel., 2018

work page 2018
[4]

Estimating depth from monocular images as classification using deep fully convolutional residual networks

Yuanzhouhan Cao, Zifeng Wu, and Chunhua Shen. Estimating depth from monocular images as classification using deep fully convolutional residual networks. IEEE Trans. Circuits Syst. Video Technol. , 2017

work page 2017
[5]

Encoder-decoder with atrous separable convolution for semantic image segmentation

Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proc. Eur. Conf. Comp. Vis., 2018 a

work page 2018
[6]

Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks

Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and Andrew Rabinovich. Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. In Proc. Int. Conf. Mach. Learn., 2018 b

work page 2018
[7]

Deformable convolutional networks

Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. In Proc. IEEE Int. Conf. Comp. Vis., pp.\ 764--773, 2017

work page 2017
[8]

Multi-task learning with multi-view attention for answer selection and knowledge base question answering

Yang Deng, Yuexiang Xie, Yaliang Li, Min Yang, Nan Du, Wei Fan, Kai Lei, and Ying Shen. Multi-task learning with multi-view attention for answer selection and knowledge base question answering. In Proc. AAAI Conf. on Arti. Intel., 2019

work page 2019

Show all 47 references

[9]

Low resource dependency parsing: Cross-lingual parameter sharing in a neural network parser

Long Duong, Trevor Cohn, Steven Bird, and Paul Cook. Low resource dependency parsing: Cross-lingual parameter sharing in a neural network parser. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conferen...

2015
[10]

Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture

David Eigen and Rob Fergus. Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. In Proc. IEEE Int. Conf. Comp. Vis., pp.\ 2650--2658, 2015

2015
[11]

Deep ordinal regression network for monocular depth estimation

Huan Fu, Mingming Gong, Chaohui Wang, Kayhan Batmanghelich, and Dacheng Tao. Deep ordinal regression network for monocular depth estimation. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 2002--2011, 2018

2002
[12]

Dynamic task prioritization for multitask learning

Michelle Guo, Albert Haque, De-An Huang, Serena Yeung, and Li Fei-Fei. Dynamic task prioritization for multitask learning. In Proc. Eur. Conf. Comp. Vis., pp.\ 270--287, 2018

2018
[13]

Understanding real world indoor scenes with synthetic data

Ankur Handa, Viorica Patraucean, Vijay Badrinarayanan, Simon Stent, and Roberto Cipolla. Understanding real world indoor scenes with synthetic data. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 4077--4085, 2016

2016
[14]

Deep residual learning for image recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 770--778, 2016 a

2016
[15]

Deep residual learning for image recognition

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 770--778, 2016 b

2016
[16]

Weinberger

Gao Huang, Zhuang Liu, Laurens van der Maaten, and Kilian Q. Weinberger. Densely connected convolutional networks. pp.\ 2261--2269, 2017

2017
[17]

Multi-task learning using uncertainty to weigh losses for scene geometry and semantics

Alex Kendall, Yarin Gal, and Roberto Cipolla. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 7482--7491, 2018

2018
[18]

Ubernet: Training a universal convolutional neural network for low-, mid-, and high-level vision using diverse datasets and limited memory

Iasonas Kokkinos. Ubernet: Training a universal convolutional neural network for low-, mid-, and high-level vision using diverse datasets and limited memory. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 5454--5463, 2017

2017
[19]

Deep attention-based classification network for robust depth prediction

Ruibo Li, Ke Xian, Chunhua Shen, Zhiguo Cao, Hao Lu, and Lingxiao Hang. Deep attention-based classification network for robust depth prediction. arXiv preprint arXiv:1807.03959, 2018

2018 arXiv
[20]

Learning depth from single monocular images using deep convolutional neural fields

Fayao Liu, Chunhua Shen, Guosheng Lin, and Ian D Reid. Learning depth from single monocular images using deep convolutional neural fields. IEEE Trans. Pattern Anal. Mach. Intell. , 38 0 (10): 0 2024--2039, 2016

2024
[21]

Hierarchical representations for efficient architecture search

Hanxiao Liu, Karen Simonyan, Oriol Vinyals, Chrisantha Fernando, and Koray Kavukcuoglu. Hierarchical representations for efficient architecture search. In Proc. Int. Conf. Learn. Repren., 2018

2018
[22]

Structured knowledge distillation for semantic segmentation

Yifan Liu, Ke Chen, Chris Liu, Zengchang Qin, Zhenbo Luo, and Jingdong Wang. Structured knowledge distillation for semantic segmentation. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., 2019

2019
[23]

Learning multiple tasks with multilinear relationship networks

Mingsheng Long, Zhangjie Cao, Jianmin Wang, and S Yu Philip. Learning multiple tasks with multilinear relationship networks. In Proc. Adv. Neural Inf. Process. Syst., pp.\ 1594--1603, 2017

2017
[24]

Cross-stitch networks for multi-task learning

Ishan Misra, Abhinav Shrivastava, Abhinav Gupta, and Martial Hebert. Cross-stitch networks for multi-task learning. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 3994--4003, 2016

2016
[25]

Revisiting multi-task learning with rock: a deep residual auxiliary block for visual detection

Taylor Mordan, Nicolas THOME, Gilles Henaff, and Matthieu Cord. Revisiting multi-task learning with rock: a deep residual auxiliary block for visual detection. In Proc. Adv. Neural Inf. Process. Syst., pp.\ 1310--1322, 2018

2018
[26]

Joint semantic segmentation and depth estimation with deep convolutional networks

Arsalan Mousavian, Hamed Pirsiavash, and Jana Ko s eck \'a . Joint semantic segmentation and depth estimation with deep convolutional networks. In 2016 Fourth International Conference on 3D Vision (3DV), pp.\ 611--619. IEEE, 2016

2016
[27]

Vladimir Nekrasov, Thanuja Dharmasiri, Andrew Spek, Tom Drummond, Chunhua Shen, and Ian D. Reid. Real-time joint semantic segmentation and depth estimation using asymmetric annotations. Proc. Int. Conf. on Robotics and Automation, 2018 a

2018
[28]

Light-weight refinenet for real-time semantic segmentation

Vladimir Nekrasov, Chunhua Shen, and Ian Reid. Light-weight refinenet for real-time semantic segmentation. arXiv preprint arXiv:1810.03272, 2018 b

2018 arXiv
[29]

Fast neural architecture search of compact semantic segmentation models via auxiliary cells

Vladimir Nekrasov, Hao Chen, Chunhua Shen, and Ian Reid. Fast neural architecture search of compact semantic segmentation models via auxiliary cells. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., 2019

2019
[30]

Extragradient method in optimization: Convergence and complexity

Trong Phong Nguyen, Edouard Pauwels, Emile Richard, and Bruce W Suter. Extragradient method in optimization: Convergence and complexity. Journal of Optimization Theory and Applications, 176 0 (1): 0 137--162, 2018

2018
[31]

Mutual learning to adapt for joint human parsing and pose estimation

Xuecheng Nie, Jiashi Feng, and Shuicheng Yan. Mutual learning to adapt for joint human parsing and pose estimation. In Proc. Eur. Conf. Comp. Vis., pp.\ 502--517, 2018

2018
[32]

Efficient neural architecture search via parameter sharing

Hieu Pham, Melody Y Guan, Barret Zoph, Quoc V Le, and Jeff Dean. Efficient neural architecture search via parameter sharing. In Proc. Int. Conf. Mach. Learn., 2018

2018
[33]

Mobilenetv2: Inverted residuals and linear bottlenecks

Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., 2018

2018
[34]

A hierarchical multi-task approach for learning embeddings from semantic tasks

Victor Sanh, Thomas Wolf, and Sebastian Ruder. A hierarchical multi-task approach for learning embeddings from semantic tasks. In Proc. AAAI Conf. on Arti. Intel., 2019

2019
[35]

Proximal policy optimization algorithms

John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

2017 arXiv
[36]

Multi-task learning as multi-objective optimization

Ozan Sener and Vladlen Koltun. Multi-task learning as multi-objective optimization. In Proc. Adv. Neural Inf. Process. Syst., pp.\ 527--538, 2018

2018
[37]

Indoor segmentation and support inference from rgbd images

Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor segmentation and support inference from rgbd images. In Proc. Eur. Conf. Comp. Vis., pp.\ 746--760, 2012

2012
[38]

Sun rgb-d: A rgb-d scene understanding benchmark suite

Shuran Song, Samuel P Lichtenberg, and Jianxiong Xiao. Sun rgb-d: A rgb-d scene understanding benchmark suite. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 567--576, 2015

2015
[39]

Rethinking the inception architecture for computer vision

Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, and Zbigniew Wojna. Rethinking the inception architecture for computer vision. Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 2818--2826, 2016

2016
[40]

Training deeper convolutional networks with deep supervision

Liwei Wang, Chen-Yu Lee, Zhuowen Tu, and Svetlana Lazebnik. Training deeper convolutional networks with deep supervision. arXiv preprint arXiv:1505.02496, 2015

2015 arXiv
[41]

Pad-net: Multi-tasks guided prediction-and-distillation network for simultaneous depth estimation and scene parsing

Dan Xu, Wanli Ouyang, Xiaogang Wang, and Nicu Sebe. Pad-net: Multi-tasks guided prediction-and-distillation network for simultaneous depth estimation and scene parsing. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 675--684, 2018

2018
[42]

Deep multi-task representation learning: A tensor factorisation approach

Yongxin Yang and Timothy Hospedales. Deep multi-task representation learning: A tensor factorisation approach. In Proc. Int. Conf. Learn. Repren., 2017

2017
[43]

Trace norm regularised deep multi-task learning

Yongxin Yang and Timothy M Hospedales. Trace norm regularised deep multi-task learning. arXiv preprint arXiv:1606.04038, 2016

2016 arXiv
[44]

Ocnet: Object context network for scene parsing

Yuhui Yuan and Jingdong Wang. Ocnet: Object context network for scene parsing. arXiv preprint arXiv:1809.00916, 2018

2018 arXiv
[45]

Joint task-recursive learning for semantic segmentation and depth estimation

Zhenyu Zhang, Zhen Cui, Chunyan Xu, Zequn Jie, Xiang Li, and Jian Yang. Joint task-recursive learning for semantic segmentation and depth estimation. In Proc. Eur. Conf. Comp. Vis., pp.\ 235--251, 2018

2018
[46]

Pyramid scene parsing network

Hengshuang Zhao, Jianping Shi, Xiaojuan Qi, Xiaogang Wang, and Jiaya Jia. Pyramid scene parsing network. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 2881--2890, 2017

2017
[47]

Practical block-wise neural network architecture generation

Zhao Zhong, Junjie Yan, Wei Wu, Jing Shao, and Cheng-Lin Liu. Practical block-wise neural network architecture generation. In Proc. IEEE Conf. Comp. Vis. Patt. Recogn., pp.\ 2423--2432, 2018

2018

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

[2] [2]

A bayesian/information theoretic model of learning to learn via multiple task sampling

Jonathan Baxter. A bayesian/information theoretic model of learning to learn via multiple task sampling. Machine learning, 28 0 (1): 0 7--39, 1997

work page 1997

[3] [3]

Efficient architecture search by network transformation

Han Cai, Tianyao Chen, Weinan Zhang, Yong Yu, and Jun Wang. Efficient architecture search by network transformation. In Proc. AAAI Conf. on Arti. Intel., 2018

work page 2018

[4] [4]

Estimating depth from monocular images as classification using deep fully convolutional residual networks

Yuanzhouhan Cao, Zifeng Wu, and Chunhua Shen. Estimating depth from monocular images as classification using deep fully convolutional residual networks. IEEE Trans. Circuits Syst. Video Technol. , 2017

work page 2017

[5] [5]

Encoder-decoder with atrous separable convolution for semantic image segmentation

Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder-decoder with atrous separable convolution for semantic image segmentation. In Proc. Eur. Conf. Comp. Vis., 2018 a

work page 2018

[6] [6]

Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks

Zhao Chen, Vijay Badrinarayanan, Chen-Yu Lee, and Andrew Rabinovich. Gradnorm: Gradient normalization for adaptive loss balancing in deep multitask networks. In Proc. Int. Conf. Mach. Learn., 2018 b

work page 2018

[7] [7]

Deformable convolutional networks

Jifeng Dai, Haozhi Qi, Yuwen Xiong, Yi Li, Guodong Zhang, Han Hu, and Yichen Wei. Deformable convolutional networks. In Proc. IEEE Int. Conf. Comp. Vis., pp.\ 764--773, 2017

work page 2017

[8] [8]

Multi-task learning with multi-view attention for answer selection and knowledge base question answering

Yang Deng, Yuexiang Xie, Yaliang Li, Min Yang, Nan Du, Wei Fan, Kai Lei, and Ying Shen. Multi-task learning with multi-view attention for answer selection and knowledge base question answering. In Proc. AAAI Conf. on Arti. Intel., 2019

work page 2019