No abstract

Machine learning (ML) solutions are prevalent. However, many challenges exist in making these solutions businessgrade. One major challenge is to ensure that the ML solution provides its expected business value. In order to do that, one has to bridge the gap between the way ML model performance is measured and the solution requirements. In previous work (Barash et al. 2019) we demonstrated the effectiveness of utilizing feature models in bridging this gap. Whereas ML performance metrics, such as the accuracy or F1-score of a classifier, typically measure the average ML performance, feature models shed light on explainable data slices that are too far from that average, and therefore might indicate unsatisfied requirements. For example, the overall accuracy of a bank text terms classifier may be very high, say 98% ± 2%, yet it might perform poorly for terms that include short descriptions and originate from commercial accounts. A business requirement, which may be implicit in the training data, may be to perform well regardless of the type of account and length of the description. Therefore, the under-performing data slice that includes short descriptions and commercial accounts suggests poorly-met requirements. In this paper we show the feasibility of automatically extracting feature models that result in explainable data slices over which the ML solution under-performs. Our novel technique, IBM FreaAI aka FreaAI, extracts such slices from structured ML test data or any other labeled data. We demonstrate that FreaAI can automatically produce explainable and statistically-significant data slices over seven open datasets.

We introduce a novel LLM based solution design approach that utilizes combinatorial optimization and sampling. Specifically, a set of factors that influence the quality of the solution are identified. They typically include factors that represent prompt types, LLM inputs alternatives, and parameters governing the generation and design alternatives. Identifying the factors that govern the LLM solution quality enables the infusion of subject matter expert knowledge. Next, a set of interactions between the factors are defined and combinatorial optimization is used to create a small subset P that ensures all desired interactions occur in P . Each element p ∈ P is then developed into an appropriate benchmark. Applying the alternative solutions on each combination, p ∈ P and evaluating the results facilitate the design of a high quality LLM solution pipeline. The approach is especially applicable when the design and evaluation of each benchmark in P is time-consuming and involves manual steps and human evaluation. Given its efficiency the approach can also be used as a baseline to compare and validate an autoML approach that searches over the factors governing the solution.

A trained ML model is deployed on another 'test' dataset where target feature values (labels) are unknown. Drift is distribution change between the training and deployment data, which is concerning if model performance changes. For a cat/dog image classifier, for instance, drift during deployment could be rabbit images (new class) or cat/dog images with changed characteristics (change in distribution). We wish to detect these changes but can't measure accuracy without deployment data labels. We instead detect drift indirectly by nonparametrically testing the distribution of model prediction confidence for changes. This generalizes our method and sidesteps domain-specific feature representation.We address important statistical issues, particularly Type-1 error control in sequential testing, using Change Point Models (CPMs,[13]). We also use nonparametric outlier methods to show the user suspicious observations for model diagnosis, since the before/after change confidence distributions overlap significantly. In experiments to demonstrate robustness, we train on a subset of MNIST digit classes, then insert drift (e.g., unseen digit class) in deployment data in various settings (gradual/sudden changes in the drift proportion). A novel loss function is introduced to compare the performance (detection delay, Type-1 and 2 errors) of a drift detector under different levels of drift class contamination.

The growing gap between the advanced capabilities of static compilers as reflected in benchmarking results and the actual performance that users experience in real-life scenarios makes client-side dynamic optimization technologies imperative to the domain of static languages. Dynamic optimization of software distributed in the form of a platform-agnostic Intermediate-Representation (IR) has been very successful in the domain of managed languages, greatly improving upon interpreted code, especially when online profiling is used. However, can such feedback-directed IR-based dynamic code generation be viable in the domain of statically compiled, rather than interpreted, languages? We show that fat binaries, which combine the IR together with the statically compiled executable, can provide a practical solution for software vendors, allowing their software to be dynamically optimized without the limitation of binary-level approaches, which lack the highlevel IR of the program, and without the warm-up costs associated with the IR-only software distribution approach. We describe and evaluate the fat-binary-based runtime compilation approach using SPECint2006, demonstrating that the overheads it incurs are low enough to be successfully surmounted by dynamic optimization. Building on Java JIT technologies, our results already improve upon common real-world usage scenarios, including very small workloads.