In spite of finite dimension ReLU neural networks being a consistent factor behind recent deep learning successes, a theory of feature learning in these models remains elusive. Currently, insightful theories still rely on assumptions including the linearity of the network computations, unstructured input data and architectural constraints such as infinite width or a single hidden layer. To begin to address this gap we establish an equivalence between ReLU networks and Gated Deep Linear Networks, and use their greater tractability to derive dynamics of learning. We then consider multiple variants of a core task reminiscent of multi-task learning or contextual control which requires both feature learning and nonlinearity. We make explicit that, for these tasks, the ReLU networks possess an inductive bias towards latent representations which are *not* strictly modular or disentangled but are still highly structured and reusable between contexts. This effect is amplified with the addition of more contexts and hidden layers. Thus, we take a step towards a theory of feature learning in finite ReLU networks and shed light on how structured mixed-selective latent representations can emerge due to a bias for node-reuse and learning speed.

Often, applications of self-supervised learning to 3D medical data opt to use 3D variants of successful 2D network architectures. Although promising approaches, they are significantly more computationally demanding to train, and thus reduce the widespread applicability of these methods away from those with modest computational resources. Thus, in this paper, we aim to improve standard 2D SSL algorithms by modelling the inherent 3D nature of these datasets implicitly. We propose two variants that build upon a strong baseline model and show that both of these variants often outperform the baseline in a variety of downstream tasks. Importantly, in contrast to previous works in both 2D and 3D approaches for 3D medical data, both of our proposals introduce negligible additional overhead in terms of parameter complexity. Although data loading overhead increases over the baseline SimCLR model (which we can show can be somewhat mitigated through parallelisation), our proposed models are still significantly more efficient than previous approaches based on sequence modelling. Overall, our proposed methods help improve the democratisation of these approaches for medical applications. • Introduce 2 models that implicitly model 3D pathology data and improve performance. • Conduct detailed ablations of the hyperparameters and design choices for our methods. • A extensive evaluation covering multiple downstream datasets, tasks, and modalities.

Learning disentangled representations with variational autoencoders (VAEs) is often attributed to the regularisation component of the loss. In this work, we highlight the interaction between data and the reconstruction term of the loss as the main contributor to disentanglement in VAEs. We show that standard benchmark datasets have unintended correlations between their subjective ground-truth factors and perceived axes in the data according to typical VAE reconstruction losses. Our work exploits this relationship to provide a theory for what constitutes an adversarial dataset under a given reconstruction loss. We verify this by constructing an example dataset that prevents disentanglement in state-of-the-art frameworks while maintaining human-intuitive ground-truth factors. Finally, we re-enable disentanglement by designing an example reconstruction loss that is once again able to perceive the ground-truth factors. Our findings demonstrate the subjective nature of disentanglement and the importance of considering the interaction between the ground-truth factors, data and notably, the reconstruction loss, which is under-recognised in the literature.

While reinforcement learning has achieved remarkable successes in several domains, its real-world application is limited due to many methods failing to generalise to unfamiliar conditions. In this work, we consider the problem of generalising to new transition dynamics, corresponding to cases in which the environment's response to the agent's actions differs. For example, the gravitational force exerted on a robot depends on its mass and changes the robot's mobility. Consequently, in such cases, it is necessary to condition an agent's actions on extrinsic state information and pertinent contextual information reflecting how the environment responds. While the need for context-sensitive policies has been established, the manner in which context is incorporated architecturally has received less attention. Thus, in this work, we present an investigation into how context information should be incorporated into behaviour learning to improve generalisation. To this end, we introduce a neural network architecture, the Decision Adapter, which generates the weights of an adapter module and conditions the behaviour of an agent on the context information. We show that the Decision Adapter is a useful generalisation of a previously proposed architecture and empirically demonstrate that it results in superior generalisation performance compared to previous approaches in several environments. Beyond this, the Decision Adapter is more robust to irrelevant distractor variables than several alternative methods.

In any game, play style is a concept that describes the technique and strategy employed by a player to achieve a goal. Identifying a player's style is desirable as it can enlighten players on which approaches work better or worse in different scenarios and inform developers of the value of design decisions. In previous work, we demonstrated an unsupervised LSTM-autoencoder clustering approach for play-style identification capable of handling multidimensional variable length player trajectories. The efficacy of our model was demonstrated on both complete and partial trajectories in both a simulated and natural environment. Lastly, through state frequency analysis, the properties of each of the play styles were identified and compared. This work expands on this approach by demonstrating a process by which we utilize temporal information to identify the decision boundaries related to particular clusters. Additionally, we demonstrate further robustness by applying the same techniques toMiniDungeons, another popular domain for player modeling research. Finally, we also propose approaches for determining mean play-style examples suitable for describing general play-style behaviors and for determining the correct number of represented play-styles.

In any game, play-style is a concept that describes the technique and strategy employed by a player to achieve a goal. Being able to identify the play-style of a player is desirable as it can enlighten players on which approaches work better or worse in different scenarios, as well as inform developers of the value of design decisions. In this paper, we propose a novel approach to play-style identification based on an unsupervised LSTM-autoencoder clustering approach for multi-dimensional trajectory-based data of variable length. We evaluate our approach on two domains and show that not only is our model capable of identifying these play-styles from entire trajectories but it is also capable of this during gameplay from partial trajectories. Additionally, it is demonstrated through state frequency analysis that the properties of each of the play-styles can be identified and compared. Through these processes, we can extract useful information which describes the different behaviours or play-styles present within a domain useful to both players and developers.

Action prediction in video games is the process of extracting useful information in order to predict the future actions of a player. Long-range dependencies and the dynamic nature of video games make it difficult for most algorithms to accurately predict the future actions of players. We propose a novel machine learning approach to improving future action prediction from video game trajectories. This method requires having first clustered player trajectories based on behaviour similarities. Our model consists of a set of LSTM based prediction modules each trained on a subset of data based upon a respective cluster. The effectiveness of our model is analysed on both a synthetic and natural dataset. We find that our future action prediction approach of leveraging multiple models trained on individual data subsets results in greater accuracy over a single model on a complete dataset.

To perform contingent teaching and be responsive to students' needs during class, lecturers must be able to quickly assess the state of their audience. While effective teachers are able to gauge easily the affective state of the students, as class sizes grow this becomes increasingly difficult and less precise. The Wits Intelligent Teaching System (WITS) aims to assist lecturers with real-time feedback regarding student affect. The focus is primarily on recognising engagement or lack thereof. Student engagement is labelled based on behaviour and postures that are common to classroom settings. These proxies are then used in an observational checklist to construct a dataset of engagement upon which a CNN based on AlexNet is successfully trained and which significantly outperforms a Support Vector Machine approach. The deep learning approach provides satisfactory results on a challenging, real-world dataset with significant occlusion, lighting and resolution constraints.

In the last few decades, much research has focused on the evaluation and assessment of students' knowledge. The idea that computers can now be used to aid assessment is appealing. While implementing automatic marking of multiple choice questions is trivial, most educators agree that such form of assessment provides only limited insight into students' knowledge. Due to this limitation, teachers prefer to use unstructured questions in assessments. These questions, however, are much more complicated to mark as the semantic meaning of a response is more important than any of the individual keywords. Latent Semantic Analysis (LSA) has a remarkable ability to infer meaning from a text in this way. This paper describes the design, implementation and evaluation of an automatic marking system based on LSA and designed to grade paragraph responses to exam questions. In addition to presenting the algorithm and the theoretical basis of the system, we describe the tests that were conducted to test its efficacy. The tests included comparing the marks for several computer science courses exams generated by the system with the original grades awarded by a human examiner. The various settings of the system are studied to understand their effect on the accuracy. Using this understanding, along with trial and error, good configurations for each question are found. Under ideal configurations the system is capable of generating marks with correlations above 0.80 compared to the human examiner's grades. This is considered an acceptable variance. Generating these ideal configurations is nontrivial but possible by designing effective ways to extract appropriate settings from features of the data. If there is enough training data the system easily performs at rates that match the inter-correlation between human markers.