Offline imitation learning (IL) refers to learning expert behavior solely from demonstrations, without any additional interaction with the environment. Despite significant advances in offline IL, existing techniques find it challenging to learn policies for long-horizon tasks and require significant re-training when task specifications change. Towards addressing these limitations, we present GO-DICE an offline IL technique for goal-conditioned long-horizon sequential tasks. GO-DICE discerns a hierarchy of sub-tasks from demonstrations and uses these to learn separate policies for sub-task transitions and action execution, respectively; this hierarchical policy learning facilitates long-horizon reasoning.Inspired by the expansive DICE-family of techniques, policy learning at both the levels transpires within the space of stationary distributions. Further, both policies are learnt with goal conditioning to minimize need for retraining when task goals change. Experimental results substantiate that GO-DICE outperforms recent baselines, as evidenced by a marked improvement in the completion rate of increasingly challenging pick-and-place Mujoco robotic tasks. GO-DICE is also capable of leveraging imperfect demonstration and partial task segmentation when available, both of which boost task performance relative to learning from expert demonstrations alone.

Parameter-Efficient Fine-Tuning (PEFT) has become the standard for customising Foundation Models (FMs) to user-specific downstream tasks. However, typical PEFT methods require storing multiple task-specific adapters, creating scalability issues as these adapters must be housed and run at the FM server. Traditional prompt tuning offers a potential solution by customising them through task-specific input prefixes, but it under-performs compared to other PEFT methods like LoRA. To address this gap, we propose Low-Rank Prompt Adaptation (LoPA), a prompt-tuning-based approach that performs on par with state-of-the-art PEFT methods and full fine-tuning while being more parameter-efficient and not requiring a server-based adapter. LoPA generates soft prompts by balancing between sharing task-specific information across instances and customization for each instance. It uses a low-rank decomposition of the soft-prompt component encoded for each instance to achieve parameter efficiency. We provide a comprehensive evaluation on multiple natural language understanding and code generation and understanding tasks across a wide range of foundation models with varying sizes.

It is well understood from literature that the performance of a machine learning (ML) model is upper bounded by the quality of the data. While researchers and practitioners have focused on improving the quality of models (such as neural architecture search and automated feature selection), there are limited efforts towards improving the data quality. One of the crucial requirements before consuming datasets for any application is to understand the dataset at hand and failure to do so can result in inaccurate analytics and unreliable decisions. Assessing the quality of the data across intelligently designed metrics and developing corresponding transformation operations to address the quality gaps helps to reduce the effort of a data scientist for iterative debugging of the ML pipeline to improve model performance. This tutorial highlights the importance of analysing data quality in terms of its value for machine learning applications. This tutorial surveys all the important data quality related approaches discussed in literature, focusing on the intuition behind them, highlighting their strengths and similarities, and illustrates their applicability to real-world problems. Finally we will discuss the interesting work IBM Research is doing in this space.

We propose a novel concept of Deep Part Embeddings (DPEs), which can be used to learn new Convolutional Neural Networks (CNNs) for different classes. We define DPE as a neuron of a trained CNN along with its network of filter activations that is interpretable as a part of a class that the neuron contributes to. Given a new class C, we explore the idea of combining different DPEs that intuitively constitute C, from trained CNNs (not on C), into a network that learns the class C with few training samples. An important application of our proposed framework is the ability to modify a CNN trained on n classes to learn a new class with limited training data without significantly affecting its performance on the n classes. We visually illustrate the different network architectures and extensively evaluate their performance against the baselines.

In this paper, we propose the Radial Loss which utilizes category and sub-category labels to learn an order-preserving fine-grained video similarity metric. We propose an end-to-end quadlet-based Convolutional Neural Network (CNN) combined with Long Short-term Memory (LSTM) Unit to model video similarities by learning the pairwise distance relationships between samples in a quadlet generated using the category and sub-category labels. We showcase two novel applications of learning a video similarity metric - (i) fine-grained video retrieval, (ii) fine-grained event detection, along with simultaneous shot boundary detection, and correspondingly show promising results against those of the baselines on two new fine-grained video datasets.

Convolutional neural networks are shown to achieve breakthrough performance for the task of single image super resolution (SISR) for natural images. These state-of-the-art (SOA) networks have been adapted to the task of single text image super resolution and have been shown to boost the optical character recognition (OCR) performance. However, these approaches depend on variations of the standard mean squared error (MSE) loss in order to train the SR network for improving the text image quality which does not guarantee optimal OCR performance. In this paper, we propose to combine the OCR performance into the loss function during network training. This results in the generation of high resolution text images that achieve high OCR performance that is comparable to the ground truth high-resolution text images and surpassing those of the SOA baseline results. We define novel intuitive metrics to capture the improvement in the OCR performance and provide extensive experiments to qualitatively and quantitatively assess improvement in the results of our proposed approach against the SOA baselines on the standard UNLV dataset.

Critical events in videos amount to the set of frames where the user attention is heightened. Such events are usually fine-grained activities and do not necessarily have defined shot boundaries. Traditional approaches to the task of Shot Boundary Detection (SBD) in videos perform frame-level classification to obtain shot boundaries and fail to identify the critical shots in the video. We model the problem of identifying critical frames and shot boundaries in a video as learning an image frame similarity metric where the distance relationships between different types of video frames are modeled. We propose a novel pentuplet loss to learn the frame image similarity metric through a pentuplet-based deep learning framework. We showcase the results of our proposed framework on soccer highlight videos against state-of-the-art baselines and significantly outperform them for the task of shot boundary detection. The proposed framework shows promising results for the task of critical frame detection against human annotations on soccer highlight videos.

Text is the easiest means to record information but need not always be the best means for understanding a concept. In psychological theories, it is argued that when information is presented visually, it provides a better means to understand a concept. While techniques exist for generating text from a given image, the inverse problem that is to automatically fetch coherent images to represent a given set of instructions (sequence of text), is a hard one. In this paper, we present a novel multistage framework to convert textual instructions into coherent visual descriptions (text instructions annotated with images). The key components in the proposed approach are: (i) novel framework, which combines the text as well as image analysis to generate visual descriptions; (ii) ensure coherency across visual descriptions, using a combination of deep learning and graph based approach. Effectiveness of our proposed approach is shown through a user study on a dataset of instructions and corresponding images collected from WikiHow website.