Pensieve (Mao, Netravali, Alizadeh — ACM SIGCOMM 2017) was the first widely-cited demonstration of reinforcement-learning-based ABR. The model takes recent throughput, current buffer level, last chosen rendition, and remaining content as input, and outputs a probability distribution over the next rendition. Training uses A3C (Asynchronous Advantage Actor-Critic) on simulated network traces drawn from real cellular and broadband measurements. The trained model outperformed BOLA and MPC on the tested traces.
Pensieve mattered as a proof-of-concept: it showed that an ABR algorithm could be learned end-to-end without explicit modeling assumptions, and could outperform careful hand-tuned heuristics. Several followup papers extended the approach to per-content models, edge-side ABR decisions, and federated training. Production deployment is rare — most operators prefer the interpretability and predictability of heuristic ABRs.
The legacy of Pensieve is broader than its direct use. The reinforcement-learning approach influenced thinking about ABR at every major streaming team. Bitmovin, THEO, and several smaller player vendors offer "ML-tuned" ABR options that draw on Pensieve-style training. Netflix's per-title ABR — which adjusts the ABR strategy per piece of content — has Pensieve as an intellectual antecedent.
The book · Volume 3 of 7
Video Streaming: A Complete Guide to Video Delivery: From ABR and CDNs to Players, WebRTC, DRM, and the Economics of Streaming
Pensieve showed that a reinforcement-learning ABR policy can beat hand-tuned rules, yet production players still rarely ship it because training traces drift from real networks. This volume covers Pensieve and the wider ABR family in detail, with buffer-based, throughput-based and hybrid rung selection, bitrate ladders, stall recovery and QoE measurement. Written by Nikolay Sapunov, CEO at Fora Soft.
Considering a learned ABR like Pensieve for your player?
Share your player telemetry and we will assess whether a learned ABR beats your current logic, with a test plan and the QoE gains to expect. 250+ video products shipped since 2005.
